Enterprise Agreement negotiations with Microsoft have always been shaped by two forces: numbers and narrative.  

Numbers reflect the economic scale of the relationship, and include metrics such as how many users are licensed, how quickly the environment is growing, and the overall size of the deal. Narrative reflects how Microsoft interprets where your organization is heading within the broader Microsoft ecosystem. 

Both have always mattered. But as Microsoft’s portfolio has expanded from individual products into an integrated ecosystem of platforms, the way Microsoft evaluates customers has become more holistic and more narrative driven.  

Today, negotiations are influenced not just by the scale of the agreement, but by the signals Microsoft interprets from how your organization adopts its technology. These signals collectively form the lens through which Microsoft determines your strategic value as a customer, and ultimately how it approaches the negotiation. Understanding this Microsoft lens is the first step in shaping your negotiation strategy. 

The signals Microsoft interprets generally fall into two categories, namely technology signals and interpretive signals. 

Technology Signals 

Technology signals reflect what you are adopting. 

Strategic Workload Adoption: Signals whether the organization is adopting Microsoft’s priority capabilities. 

Examples: 

• Copilot(s) 
• Power Platform 
• Advanced Security Workloads: Entra Suite, Advanced Purview, Sentinel 

These indicate whether your organization is moving toward Microsoft’s strategic workloads. 

Platform Depth: Signals how deeply the organization has committed to the M365 platform. 

Examples: 

• Microsoft 365 suite level (E3 → E5 → E7) 
• Frontline worker expansion 
• Security portfolio rationalization (Microsoft vs. third-party tools) 

These signals indicate whether the M365 platform is becoming foundational within the organization’s environment. 

Ecosystem Expansion: Signals whether Microsoft’s footprint is expanding beyond the core M365 platform. 

Examples: 

• Azure workload adoption 
• Dynamics deployments 
• Unified Support agreements 

These indicate whether the organization is moving toward broader adoption of the Microsoft ecosystem 

Interpretive Signals 

Interpretive signals reflect how Microsoft interprets your strategic trajectory relative to their overall vision. 

Deployment Momentum: Signals how quickly new capabilities move from evaluation to production deployment. 

Examples: 

• Copilot pilots expanding into enterprise rollout 
• Defender deployments replacing existing security tools 
• Automation capabilities scaling across departments 

Momentum signals whether platform adoption is likely to accelerate over time. 

Strategic Influence: Signals your potential value in influencing other customers. 

Examples: 

• Iconic brand 
• Reference customer potential 
• Reference case studies for strategic workloads 

These signals can influence Microsoft’s willingness to invest strategically in the relationship. 

Traditional Metrics: Signals related to the economic scale of the agreement. 

Examples: 

• Deal size 
• Environment growth trajectory 
• Renewal timing 

Traditional metrics still matter, but they now represent only one category of signals Microsoft interprets when evaluating a customer. 

The Amplifying Effect of Specificity 

All of these signals become more powerful when they are specific and actionable. An abstract signal says: “We plan to evaluate Microsoft security,” whereas a specific signal says: “Our Okta, SailPoint, and CrowdStrike contracts expire within the next 12 months, and we intend to evaluate whether Microsoft can replace those platforms.” 

Specific signals are easier for the Microsoft account team to communicate internally and easier for the business desk to evaluate. As negotiations lean more heavily toward narrative, clear and specific signals carry greater influence than abstract intent. 

It is also important to recognize that customers rarely interact directly with the final commercial decision makers inside Microsoft. Pricing concessions and investment decisions typically flow through the business desk, which evaluates deals based on the narrative carried internally by the account team. The clearer the signals are, the easier it is for that narrative to be communicated internally. 

Together, these signals form the internal narrative Microsoft builds about your organization. That narrative shapes how Microsoft evaluates you and ultimately influences the negotiating posture you encounter as you approach your next EA renewal. 

The E7 Wildcard 

The framework in the previous section helps explain how Microsoft interprets signals about a customer’s trajectory within the ecosystem. But it assumes something important: that the structure of the platform itself remains relatively stable. 

For years, the Microsoft 365 platform had a predictable ceiling. Most enterprise customers sat somewhere along a spectrum that ran from Microsoft 365 E3 to E5, with negotiating leverage shaped by how early or deeply an organization aligned with the top licensing tier.  

Microsoft 365 E7 introduces a new variable. By adding a strategic tier above E5, Microsoft has effectively raised the ceiling for platform alignment. When the top of the platform stack moves upward, the leverage bands beneath it shift downward. What previously represented the leading edge of platform alignment may now sit one step lower on the curve. 

Organizations that adopt the new E7 tier will sit at the front edge of the adoption lifecycle, where negotiating leverage is strongest. Customers positioned just behind that tier running a broad E5 ecosystem may still have strong leverage, but they no longer represent the leading edge. Organizations further down the curve may find that little negotiating leverage remains if they choose not to move up the stack. 

How do I put this to use? 

By this point, you’ve seen how Microsoft evaluates you: through a combination of signals, interpreted through the lens of the Microsoft 365 platform and the broader ecosystem. You’ve also seen how the introduction of E7 will shift the leverage landscape. But understanding Microsoft’s lens is only half of the equation. 

The lens that ultimately determines whether a negotiation succeeds is your own. Negotiation leverage ultimately emerges from the gap between your roadmap and Microsoft’s. Microsoft is evaluating where it believes your organization is heading within its ecosystem. You need to evaluate where your organization actually needs to go. 

In the next article, we’ll shift the perspective and look at the negotiation through your lens: how to evaluate the long-term value of Microsoft 365 across your environment, and how to build a roadmap that positions you for your next EA renewal. 

In late Feb. 2026, Microsoft added a notice in the Microsoft 365 Admin Center (MC949965) about a delay in the rollout of its New Outlook for Windows.  

“The opt-out phase for Enterprise environments will now begin in March 2027 (previously April 2026), providing organizations with 12 months of lead time to prepare.” 

(Microsoft’s update noted that the GCC High and DoD rollout timelines will be provided at some later date. 

Microsoft’s update claimed the company is seeing “strong and accelerating adoption of new Outlook.” But it added that Microsoft is pushing to expand capabilities and address “feedback from customers who want to go further with the new Outlook.” 

The Long, Slow March to New Outlook 

Microsoft has told customers to expect a handful of milestones as it moves to replace the current Outlook with the new Outlook.  Microsoft had designated the New Outlook as “Opt-in” and off by default in Mar. 2024. Microsoft declared New Outlook to be generally available shortly thereafter, on Aug. 1, 2024.  

New Outlook was scheduled to hit the “Opt-out”/on-by-default  milestone in April 2026. But now Opt-out isn’t slated until March 2027. After opt-out, managed customers will receive a minimum of 12-month notice before the “Cutover” stage begins — now, no earlier than March 2028. After Cutover, customers won’t be able to switch back to Classic Outlook, and new deployments of Outlook with Microsoft 365 subscriptions will come with the New Outlook.  

However, in 2024, Microsoft admitted that existing classic Outlook installations — both perpetual and subscription — will continue to be supported until at least 2029. After Directions on Microsoft asked, Microsoft said that existing enterprise customers with a Microsoft 365 subscription or license that includes desktop apps will be able to download classic Outlook separately and use for no additional fee. 

New Outlook has missing features and annoyances, the same as any Office app in beta testing. The trouble is, it’s not in beta testing. It’s generally available and it was about to land on the desktop of every Microsoft 365 Apps for enterprise user, unless their enterprise opted out.    

Many will opt out, some until 2029, the earliest date that classic Outlook could be retired. These companies now have 11 more months to adjust their Office deployment and updating processes to keep New Outlook out. 

And If You Don’t Want to Delay… 

What about companies that want New Outlook? There will be some, although they might not know it yet. Outlook already uses machine learning and other AI methods for things like spam filtering and scheduling. Generative AI is accelerating improvements, and the improvements will mostly come to New Outlook, not Classic. 

 Companies that see payoff from AI in apps like Outlook can’t cling to Classic for long. For these companies, the delay means more time to clear away migration blockers, like classic Outlook extensions that don’t work on New Outlook.  

There’s also possibly the biggest migration blocker of all: Exchange on-premises or in third-party hosting . New Outlook treats these Exchange systems  like any other Internet mail service, with no more calendar integration or other features than you would get with Yahoo or Google. That means companies that want the coming AI capabilities of New Outlook will have to move to Exchange Online. Unless, of course, Microsoft changes its plans. 

For more details on the evolution of e-mail and collaboration at Microsoft, see the Directions Business Applications Roadmap. 

Microsoft will launch a new, high-end Microsoft 365 licensing tier, Microsoft 365 E7, on May 1, priced at US$99 per user per month. For that price, customers will get all the features in Microsoft 365 E5, Microsoft 365 Copilot and Agent 365 — as well as all the network access, identity protection and governance features in the Entra Suite — in a single bundle.  

In addition to including its Agent 365 control plane in E7, Microsoft also plans to make Agent 365 available to IT and security professionals on May 1 as a separate add-on, priced at US$15 per user per month. Microsoft’s current top-of-the-line M365 licensing bundle is E5, which as of July 1, will be $60 per user per month and does not include the US $30 per user per month M365 Copilot. To date, only three percent of Microsoft’s 450 million commercial M365 customers have purchased Microsoft 365 Copilot.

Microsoft did not disclose any details about how or if agents will be licensed as part of M365 E7. Sources had said Microsoft was looking at some kind of hybrid per-user and consumption, or metered, licensing model for E7. So far, however, the company has not talked about the details for providing agents with their own Teams, OneDrive, and e-mail accounts, as Microsoft officials have indicated will be necessary for at least some agents, (We asked Microsoft for more information on the licensing for digital agents, but no word back so far.) 

Microsoft announced its E7 and Agent 365 plans on March 9, a week after Directions on Microsoft and others heard from our contacts that the new E7 bundle was coming soon. 

On March 4, CEO Satya Nadella did touch on this topic during his appearance at a Morgan Stanley tech conference. 

When it comes to “digital workers,” Nadella said, “I think the combination of subscriptions with some limits of usage plus a meter is where I think we will end up. And it doesn’t matter if it’s a person or an agent. And I think the first place where this is happening already with the business models is in coding. And I think information work also will happen. So from a TAM (total addressable market) expansive perspective, for us, I look at all agents as users and with maybe more flexibility on how people license that.” 

Agent 365 is one key to this coming subscription model. Agent 365 — which is currently in preview as part of Microsoft’s “Frontier” test program — is not a service. It’s a set of features including identity management using Entra, compliance controls using Purview, and security infrastructure using Defender XDR. Agent 365 supports agents built using Microsoft tools, third-party and open-source frameworks, and those running in partner clouds. 

Copilot Cowork: Claude Cowork Meets M365 Copilot 

On March 9, Microsoft also took the wraps off plans for a new feature coming to Microsoft 365 Copilot called “Copilot Cowork.” Copilot Cowork is built on the technology that powers Claude Cowork and was co-developed by Microsoft and Anthropic. Copilot Cowork will help customers handle “long-running, multi-step tasks,” Microsoft officials said. Copilot Cowork is currently in private pilot with select customers and will be available later in March as a research preview via Microsoft’‘s Frontier test program. 

Last fall, Anthropic announced a connector for Claude and Microsoft 365. But this new integration between the two product families doesn’t require separate connectors or integration. 

Copilot Cowork will work in conjunction with the built-in “agentic experiences” (technology Microsoft previously referred to as “Agent Mode”) in Word, Excel, PowerPoint and Outlook that also are meant to assist with multi-step tasks. Both Claude and Open AI models are available in Copilot Chat, and can work alongside Microsoft’s own dedicated Word, Excel and PowerPoint agents, which are meant to help specifically with document creation inside those applications. 

Even though Microsoft has a long-standing partnership with OpenAI to build products and services on top of OpenAI’s GPT technology, it has been hedging its AI bets by working more closely with and investing in Anthropic in recent months. Earlier this year, Anthropic made Claude Cowork available on Windows, leading some industry watchers to wonder whether Claude Cowork might prove to be more useful than Microsoft’s own M365 Copilot and agents for helping automate tasks for knowledge workers. 

Initial reports around the expected new Microsoft 365 E7 have focused on pricing and packaging, presenting it as a higher-end bundle that includes AI features and carries a higher price tag. That framing makes sense. Microsoft has long relied on expanding bundles at expanding price points. But that perspective misses the more important story. 

If the reports are directionally accurate, E7 is not simply “E5 plus AI for more money.” It represents a structural move in how Microsoft intends to anchor AI inside the enterprise: By turning it into a platform. Building scalable, enterprise-grade platforms is where Microsoft has always been strongest. 

Assuming E7 bundles Copilot and Agent 365 into a single subscription, as we’re hearing, that signals something larger than monetization. It suggests Microsoft is formalizing AI as a platform layer across its enterprise stack.  That is why E7 appears less like a new suite and more like a next-generation platform. 

The Control Plane for Digital Workers 

Microsoft is positioning AI agents alongside human users. That is why identity architecture is foundational. Entra is no longer simply directory and access management. Purview is no longer just compliance tooling. Defender is not merely a security suite. Together, Entra, Purview, and Defender form the enterprise control plane. 

AI agents operating at scale require structured identity, conditional access, telemetry, and governance guardrails. That is not optional infrastructure. It is prerequisite infrastructure. If digital workers are going to operate inside enterprise workflows, they must be governed with the same rigor as human workers. 

Microsoft has already been strengthening the governance and security components of that control plane. Purview has expanded into AI governance scenarios. Intune Suite capabilities and Security Copilot have been drawn more tightly into core bundles. The direction is clear: advanced governance and security capabilities are moving toward the center of the platform. 

If E7 is intended to represent a true enterprise-grade AI tier, Entra would logically need to elevate identity alongside Purview and Defender. In a world where humans and digital agents operate side by side, Entra, Purview, and Defender must function as a unified, advanced layer — not as mismatched tiers of capability. 

That is what a complete enterprise AI control plane would require. 

Seat Licensing Meets Consumption Economics

Early reporting also mentions the possibility of E7 using hybrid licensing, blending per-seat subscription with consumption-based components. That reinforces a trend we have been watching closely: the convergence of Microsoft 365 licensing and Azure economics. 

Traditional M365 licensing is seat-based. You pay a fixed price per user per month. If you have 10,000 employees licensed for E5, you know what that line item will be. The cost scales with headcount. 

Consumption pricing works differently. Instead of paying per user, you pay for activity. That may mean compute cycles, AI inference calls, storage processed, or data analyzed. Azure already operates this way. The meter runs when workloads run. 

We are already seeing these models converge. Power BI Pro is licensed per user, but Fabric capacity is metered. Purview capabilities inside E5 are user-based, yet extended data scanning and cross-cloud governance rely on Azure consumption. The boundary between seat licensing and metered activity is already blurring. 

AI introduces a new dynamic. Organizations have hesitated to deploy Copilot broadly as a  per-user add-on because the cost of entry is visible and immediate. But enterprise AI scale is expensive. Microsoft must monetize that reality, and Azure provides the mechanism to do so. A hybrid licensing model would enable Microsoft to turn on advanced AI capabilities broadly while charging based on activity. Turn it on. Pay for what runs. 

Microsoft 365 E7 consumption billing could take multiple forms. E7 could provide a base user entitlement, with advanced AI workloads metered through Azure, similar to Fabric and extended Purview scenarios. Or Microsoft could think even bigger. 

What if E7 shifts materially toward Azure-based metering? Imagine a model where the full AI-integrated stack is broadly enabled, but cost scales with activity rather than headcount. For years, enterprises have paid for features they rarely use. A metered model changes that equation. It lowers the barrier to adoption. It allows organizations to scale AI gradually instead of committing large upfront license expansion. It aligns cost with actual workload intensity. 

At the same time, it shifts the model toward activity-based economics, reducing shelfware risk while allowing AI scale to grow in proportion to adoption. If E7 moves in this direction, it would not simply expand Microsoft’s AI footprint. It would reshape how enterprises evaluate licensing in an AI-driven world. 

Regardless of the final structure, the signal is clear. The licensing model itself is evolving to accommodate AI at scale. Organizations should prepare to manage seat-based licensing and metered consumption as a unified strategy rather than as separate budget lines. 

E7 as Microsoft’s Next Big Bet 

For years, Microsoft has owned the core enterprise productivity surface area: identity, collaboration, endpoint management, compliance, and security. That installed base gives Microsoft deep integration into enterprise operations. If E7 formalizes AI as part of that control fabric, this becomes less about upselling E5 customers and more about securing Microsoft’s long-term role at the center of enterprise AI operations. 

The companies that win the AI era may not be those with the most advanced models. They may be those that control the operating environment in which those models run. If the early signals are accurate, E7 could represent Microsoft placing that bet.

Microsoft is expected to introduce a new Microsoft 365 subscription sometime in the next few months: The long-rumored E7. But M365 E7 isn’t simply Microsoft 365 E5 plus M365 Copilot, according to sources. The coming E7 will be key to Microsoft’s plan to license “agentic” workers like human employees, Directions on Microsoft hears. 

Like Business Insider, Directions is hearing that the E7 subscription will include Microsoft 365 Copilot, as well as the company’s recently introduced Agent 365 agent-management control plane. The E7 licensing tier also could include Entra features that are not currently part of Microsoft 365 E5 as part of a broader push to integrate digital identity and governance. 

Agent 365 — which is currently in preview as part of Microsoft’s “Frontier” test program — is not a service. It’s a set of features including identity management using Entra, compliance controls using Purview, and security infrastructure using Defender XDR. Agent 365 supports agents built using Microsoft tools, third-party and open-source frameworks, and those running in partner clouds. These agents can be built using Copilot Studio, Microsoft Foundry, the M365 Agents Toolkit, the Agent Framework, and the Agent 365 SDK.

We asked Microsoft for comment on M365 E7, but no word back so far. Update (March 3): “Microsoft does not have anything to share at this time,” a spokesperson told us.

M365 E7: First New Enterprise Subscription Since 2015

Microsoft’s decision to go ahead with an E7 subscription is surprising in some ways, and not so much in others. 

M365 E7 will be the first new M365 licensing tier for enterprises that Microsoft has introduced since it launched M365 E5 in 2015. In spite of Microsoft’s constant pressure on customers to move users to M365 E5, uptake has been gradual, though picking up steam in recent months. (In 2022, the last time Microsoft disclosed publicly an E5 subscriber count, officials said 12 percent of its Office 365/Microsoft 365 installed base were E5 subscribers.) 

In recent years, Microsoft has been introducing pricey security, governance and collaboration add-ons to M365 E5 rather than integrating these features into its existing enterprise subscriptions. Late last year, however, Microsoft announced plans to fold some previously separately priced add-ons — such as Defender for Office Plan 1’s enhanced email security features and certain advanced Intune capabilities — into its E3 and E5 plans.  

Because Microsoft 365 Copilot’s adoption to date has been modest, with just over three percent of Microsoft’s 450 million M365 business subscribers purchasing seats, Microsoft no doubt is looking for ways to try to entice more business customers to purchase and use Microsoft 365 Copilot. In recent months, the company has been discounting M365 Copilot, the list price of which remains US$30 per user per month for enterprises, in its licensing deals with some customers, but the price reduction seemingly has not moved the needle much. 

Starting July 1, Microsoft will be increasing its Office and Microsoft 365 suite prices across the board. M365 E3 will go from US$36 to US$39 per user per month. And E5 will jump to US$60 per user per month from its current US$57 price. It would make sense for Microsoft to introduce E7 around that time at a price that would make it interesting to those facing these coming price hikes. 

What About E7 Pricing? 

The biggest question for many when it comes to E7 is the price. Business Insider’s sources say the new E7 subscription could be priced at US$99 per user per month.  But E7 needs to be priced very competitively to win over those who’ve been largely fence-sitting during Microsoft’s M365 Copilot AI push.  

Microsoft officials have said to expect agents, as they proliferate in the enterprise, to need to be licensed in ways similar to human employees. They’ll need Entra IDs, email accounts, OneDrive accounts, Teams accounts, and more. If the inclusion of Agent 365 in E7 gives customers a simple and affordable way to license digital agents, it might be worth more than simply E5 (at $60 per user per month) plus M365 Copilot (at $30 per user per month) combined. 

We hear Microsoft could be planning to switch up how E7 is priced. Instead of simply going with a per-user subscription fee, Microsoft could be looking at some kind of hybrid user- and consumption-based pricing as the way to go here. 

“Potential consumption-based pricing reinforces the convergence between M365 licensing and Azure economics. That would align AI workloads more directly with Azure revenue models and deepen the M365/Azure integration story,” said Directions’ Director of Advisory Services Lane Shelton. 

Shelton added: “This isn’t about a new licensing tier. It’s about Microsoft positioning itself as the enterprise AI control plane for the emerging digital worker.” 

For years, Enterprise Agreement (EA) renewals have been high-stakes exercises. They required preparation, modeling, and disciplined negotiation. Even in their difficulty, renewals followed a familiar rhythm. Controlling costs required real work. You anticipated pricing pressure. You expected negotiation friction. Renewal season was a battle, but it was a battle you had seen before. You understood the terrain. 

Now the terrain itself has changed. 

Over the past several years — and especially since the November 2025 elimination of subscription volume discounts — the commercial structure underneath EA renewals has shifted. What used to be a gradual slope of increases is now a sharper step. The ladder customers once climbed, trading deeper platform alignment for improved leverage, has fewer rungs.  

Consider the pattern many enterprises have lived through. Imagine your EA cost $100 per year in the perpetual era. You negotiated a Full Platform EA because it offered the most leverage available. When renewal came, Microsoft introduced Office 365. You were told the Full Platform construct no longer carried the same commercial weight. You could keep it, but it would now cost $130. Or you could move into the subscription model for $115. You moved. At the next renewal, Microsoft introduced M365. O365 would now be $145, but M365 could be had for $130. You moved again.  

Later, M365 E5 became the new strategic tier. Discounts on E3 softened; E5 became the lever. You traded up. Each time, the increase felt manageable because there was something to move toward. There was always another rung on the ladder. Today, many organizations have reached the top of that ladder, and in the post–volume discount era, there may not be another rung to trade for leverage. 

The New Layers of Pressure 

All the usual negotiating challenges still exist. But over the past few years, Microsoft has layered on new ones. 

• Add-On Pressure 
  M365 E5 was positioned as the flagship — the suite that had “everything.” But over the past several years, Microsoft has increasingly introduced high-value capabilities as paid add-ons rather than bundled enhancements. Teams Premium, Intune Suite, Copilot, Entra ID Governance, Power Platform capacity, Viva modules. And the list continues to expand. Even organizations already on E5 often find that experiencing the full M365 ecosystem now requires E5 plus significant incremental spend. 

• Contract Shift 
  Agreement structure is no longer a background detail. Microsoft has been encouraging some customers toward Cloud Solution Provider (CSP), while moving others into direct Microsoft Customer Agreement for Enterprise (MCA-E) agreements, and reshaping engagement models for large enterprises. These agreement vehicles carry different pricing mechanics, partner dynamics, and usage implications. Renewing into “another EA” is no longer automatic, and structure itself has become a strategic lever. 

• Azure Convergence 
  More functionality is now monetized through Azure consumption meters rather than traditional licensing. Expanded capacity, AI workloads, security telemetry, and data governance capabilities often flow into variable, usage-based billing. These costs are harder to forecast and easier to underestimate, especially when they originate inside the M365 ecosystem but surface in Azure spend. 

None of these layers alone creates a crisis. But when you stack them on top of traditional renewal pressures, and then remove the structural volume discount tiers, as Microsoft did in November 2025, the math changes. For many organizations, especially large enterprises, it could feel like a steep change. A material shift in annual Microsoft run rate that cannot be absorbed casually.  

That is the Financial Cliff. The first step in navigating it is knowing your number. 

How do you measure your potential Financial Cliff? 

The Financial Cliff is the gap between your current annualized Microsoft run rate and your projected post-renewal run rate under the new model. Most organizations don’t know this number until late in the cycle. That’s a mistake. You cannot negotiate what you have not quantified. Here’s how to compute it: 

1. Model your current annualized run rate. 
Use your current license quantities and the pricing from your existing EA. This tells you what your renewal would look like if you could simply extend your current agreement for another cycle. 

2. Model your projected renewal run rate. 
Apply post–November 2025 pricing — and, where relevant, post–July 2026 pricing — to those same quantities. Remove past pricing concessions. Remove waterfall discount tiers. Apply any pricing protections you have, such as Not-to-Exceed clauses. 

Pro move: If you’re already evaluating major add-ons at scale, or you can quantify Azure consumption tied to M365 workloads, layer that in. But start with the structural reset first. 

3. That delta is your Financial Cliff. 
Think of it as a ledger. On the left side is what you would pay if you could preserve today’s pricing. On the right side is what you’ll likely pay at renewal under the new structure. That gap is what you are negotiating. 

How Not to Fall Off the Cliff 

Historically, customers could rely on structural levers like scale, tenure, and platform alignment to soften renewal impact. In the post–volume discount era, those levers function differently. Microsoft now evaluates customers holistically. Volume still matters, but alignment, consumption trajectory, workload adoption, and strategic posture matter more. 

There are new variables at play, but they can be turned into new leverage if applied correctly. In the posts that follow, we’ll unpack how to prepare — how Microsoft evaluates you as a customer, how M365 now anchors your leverage, how agreement strategy has become multidimensional, how Azure increasingly influences your EA, and how to position your organization for maximum leverage. 

For now, compute your Financial Cliff. Have the conversation early. Nothing else matters if you don’t start with the number. 

Microsoft is releasing two different versions of Windows this year: Windows 11 26H1 and Windows 11 26H2. For the vast majority of business customers, 26H1 is going to be a nothing- burger; it only will run on Qualcomm Snapdragon X2-based devices — plus some other non-specified “select new silicon” — that will come to market in early 2026.  

This means for the vast majority of enterprises, the one and only new Windows release that will matter in 2026 is Windows 11 26H2, which will be out later this year, most likely around Oct.  

“Organizations should continue to purchase, deploy, and manage devices running broadly released versions of Windows 11 (e.g. versions 24H2 and 25H2) with confidence,” according to guidance in a Feb. 10 Microsoft blog post. 

The post adds:  

“In short: Windows 11, version 26H1 should not impact your current Windows deployment and purchasing strategy. There is no benefit to waiting or deferring plans based on version 26H1, unless you are specifically targeting adoption of devices with silicon that requires such.” 

What Windows 11 26H1 Is … and Mostly Isn’t 

Microsoft won’t make 26H1 available to existing Windows 11 users through the usual distribution channels, as it’s not a feature update to Windows 11 25H2. There will be no 26H2 update for those who get 26H1 because Windows 11 26H1 is based on a different Windows core than Windows 11 24H2, 25H2 and the coming 26H2.  

(For those who have followed along with Microsoft’s Periodic-table-inspired Windows codenames, Windows 11 24H2, 25H2 and 26H2 are based on the “Germanium” platform release, while 26H1 is based on the release codenamed “Bromine,” as Windows Central explains.) 

Microsoft notes that devices that run 26H1 will have “a path to update in a future Windows release,” which many company watchers are assuming will be Windows 11 27H2.  Some are wondering if that release might be marketed as “Windows 12” — or maybe some other Copilot- and/or AI-themed name of the moment. 

Windows 11 26H1 will not support hotpatch updates, though it will work with Windows Autopatch, Intune and Configuration Manager, Microsoft has said. However, 26H1 will get monthly security, quality, and feature updates on a regular basis, officials said. 

Microsoft added entries for Windows 1 26H1 to its lifecycle product pages on February 10. For enterprises and education, the end of support date for 26H1 is March 13, 2029; for Home and Pro customers, the end of support for 26H1 is March 14, 2028. Additionally, the Microsoft Release Health Dashboard, which also now includes an entry for 26H1, notes that starting with Windows 11 26H1, .NET Framework 3.5 is no longer a Windows Feature on Demand optional component, and support for it ends Jan. 9, 2029. 

Going Back to Windows Basics 

Microsoft’s disclosure about cleaving Windows, at least temporarily, into two branches comes at an interesting time.  

Following months of Windows customer complaints about Microsoft prioritizing AI over operating-system fundamentals (as evidenced by a number of recent problematic Windows patches), company officials said earlier this month that they plan to get back to basics with Windows and focus on quality and reliability. 

Shortly thereafter, Microsoft announced via a blog post some changes it is making in Windows around transparency and consent. As part of this work, the company plans to add a “Windows Baseline Security Mode,” which will ensure that only properly signed apps, services, and drivers are allowed to run, with the option for users and IT admins to override these settings for specific apps when needed. Microsoft also said Windows will now prompt users when trying to access sensitive resources like files, cameras or microphones, or install other unintended software. In addition, apps and AI agents will give users and IT admins better visibility into their behaviors. All of these changes will “roll out through a phased approach,” officials said, without providing a specific timeline.  

Separately, Microsoft is starting to roll out new Secure Boot certificates for older PCs as part of the regular monthly Windows updates (or via organizations’ own update processes). The Secure Boot Certificates, stored inside PC firmware, are set to start expiring in June 2026. Secure Boot technology is meant to block untrusted code at the earliest stages of the boot process. 

Microsoft has released for the first time an AI adoption number about which many had been curious. During its Q2 FY26 earnings call on January 28, officials said the company now has 15 million paid seats of Microsoft 365 Copilot among its 450 million Microsoft 365 commercial subscribers.

This 15 million figure does not include consumer Copilot subscriptions. Nor does it include the Microsoft M365 and Office 365 subscribers using Copilot Chat who do not have an additional paid M365 Copilot subscription; Microsoft says there are “multiples more” of those enterprise Chat users. In other words, only 3.33 percent of M365/O365 commercial customers have bought Microsoft 365 Copilot licenses for $30 per user per month — or less, based on recent discounting of the service.

Microsoft 365 Copilot has been generally available to commercial customers since Nov. 2023. Even though M365 Copilot is powered by OpenAI’s GPT technology, Microsoft has begun including Anthropic’s Claude models in M365 Copilot to give users more choice and reduce its dependency on OpenAI. It also has added a couple of Microsoft-developed agents, the Researcher and Analyst agents, to M365 Copilot for no additional charge to try to sweeten the value proposition. By integrating Copilot Chat into M365 and O365 subscriptions for no additional charge, Microsoft has tried to show Office users the value of the tool. But it seems even a built-in freebie hasn’t resulted in a guaranteed mass upgrades to the paid Microsoft 365 Copilot.

“For comparison, OpenAI claimed roughly 35 million paying users (albeit laregly consumers, not enterprises) as of last summer, and they don’t have anything other than the core product to draw them in,” said Directions on Microsoft analyst Greg DeMichillie. “My conclusion is that Office isn’t much of an advantage, at least in terms of getting people to shell out money for AI. Of course, you could argue they just haven’t done a good job integrating into Office yet.” 

Microsoft has a history of relying on suites and bundles to power it through until it eventually “wins,” noted Directions‘ analyst Jim Gaynor. But “the old cash cows may no longer have the juice to cover the massive costs of AI,” Gaynor added.  

“Microsoft is struggling mightily to be a leading-edge player here and not to become the ‘utility service provider’ of infrastructure (Azure). After all, there’s no hockey-stick growth in that. But the very things that have carried them in the past may be a disadvantage now,” Gaynor said. 

About That $625 Billion Backlog… 

Unsurprisingly, Microsoft officials don’t see things that way. 

“Even in these early innings, we have built an AI business that is larger than some of our biggest franchises that took decades to build,” said Microsoft CEO Satya Nadella during the earnings call.  

Revenue for the quarter was $81.3 billion; net income, $38.5 billion; and earnings per share $4.14, ahead of expectations. Azure and related services grew 39 percent in the second fiscal quarter, compared to 40 percent in Q1. 

Microsoft is continuing to invest in supporting cloud and AI workloads at a breakneck pace. Officials said capital expenditures for the quarter were $37.5 billion, up 66 percent year-over year (and up from $34.9 billion in Q1 FY26). Roughly two-thirds of the $37.5 billion were primarily for GPUs and CPUs for Azure and AI solutions; the remainder was for “long-lived” assets like datacenter leases. Officials said that customer demand is still continuing to exceed supply, but are expecting capex spending to decrease next quarter.

Microsoft officials also acknowledged that its backlog of cloud orders more than doubled in Q2, to a whopping $625 billion. Of this total, 45 percent is attributable to OpenAI’s Azure commitments over multiple years, which worries some market watchers who have been cautious about OpenAI’s plans to become profitable. 

Another new data point shared during the call: Microsoft says it now has 1 billion Windows 11 users, up 45 percent year-over-year. The end of support for Windows 10 in Oct. 2025 fueled this growth. 

The New Microsoft Tech Stack 

Nadella outlined the three layers of its new tech stack where Microsoft is expecting growth in the years to come. Remaking GitHub Copilot into a broader, agent-centric platform is a big part of the plan. The three layers:  

• Cloud and token factory 

• Agent platform 

• Agentic experiences 

Cloud and token factory is the layer where Microsoft is working to build out the underlying cloud infrastructure powering first- and third-party apps. The company is optimizing for “tokens per watt per dollar” using silicon, systems and software, Nadella said.  He highlighted the Fairwater “AI superfactory” datacenters that Microsoft is building, as well as its own Maia 200 AI accelerator processor as areas of investment here.  

Nadella said Microsoft’s view is “all software is being rewritten,” and “agents are the new apps.” Microsoft is positioning Microsoft Foundry and Copilot Studio as its hubs for customers building apps and agents. And the company is betting on its recently introduced IQ brands across Fabric, Foundry and Microsoft 365 to connect agents to systems of record and all kinds of data, he said. He claimed that Foundry is not just an AI driver, but a driver for other Azure developer, database and app services.  

To further its agentic ambitions – and hopefully thwart those of its AI-coding competitors — Microsoft is working to turn GitHub Copilot into “an ecosystem of agents that collaborate across the entire development lifecycle from coding and code review to security, debugging, deployment, and maintenance.” 

Microsoft has big ambitions in the AI space, but its approach requires customers to change the way they work and to build agents/apps from scratch. The big questions: Will customers buy in? To what extent, and when?

It’s only January, but already 2026 is shaping up to be a critical year for Microsoft AI, the tech industry, and everyone who depends on Microsoft products.

Directions on Microsoft‘s VP of Research Barry Briggs breaks down Microsoft’s AI strategy and long-term bets in this video. He details how AI could reshape business, productivity, and the workforce, along with the opportunities and risks of this new industrial revolution.

Microsoft’s AI deliverables this year look quite a bit different than it did last year. The big question: Will Microsoft’s AI strategy pay off?

The last three years have seen Microsoft pivot to a hard focus on AI, similar to how the company turned to the web browser and internet in the late 1990s or tried to shoulder up to Apple and Google in mobile in the late 2000s. These days, it seems like every product and group in Microsoft is either finding a way to co-opt the Copilot brand, or forcibly bolting Copilot to some otherwise unrelated technology in an effort to ride the AI wave and be relevant. But unlike those earlier pivots, the AI tsunami comes with the massive expense of building out datacenter infrastructure with expensive servers, thereby forcing Microsoft to cut costs and try to increase revenue elsewhere to feed the beast. 

It’s understandable from a business strategy perspective. Microsoft views AI as an existential development and so do investors looking for the Next Big Thing to move the line and up and to the right. Microsoft’s own lore tells the company it can dominate if it just focuses enough, as it did against Netscape in the “browser wars,” and also warns of the high cost of losing out, as it did with mobile (and Nokia). 

But for the millions of customers that rely on Microsoft for stable, predictable technology, especially the businesses that view tech as a reliable tool for more traditional markets, Microsoft’s focus on AI comes as a heavy tax on the more “boring” technologies they depend on.  

This isn’t entirely new. Customers are used to being pushed in Microsoft’s strategic directions—whether it’s the transition from on-premises Office software to hosted Microsoft 365 services, or changes from perpetual licenses to subscriptions. It usually comes with a mix of carrot and stick, with Microsoft cajoling customers forward with new features or temporarily lower prices and pushing from the rear by withholding features from legacy products and reducing interoperability with newer services. 

However, the focus on AI is having a greater impact than Microsoft’s previous obsessions strategic pivots. 

The Impact of Betting the Company 

In 2025 Microsoft laid off more than 15,000 across the company. Combined with RTO mandates designed in part to encourage remote workers to resign, many teams were drastically reduced or eliminated. This has scaled back or delayed roadmap milestones on several services, and Microsoft has been even more aggressive at culling long-running “previews” and sunsetting software (just look at our Roadmaps for the growing list of deprecations and retirements.) 

Meanwhile, several long-standing products suffer from benign neglect – still supported but without new features – while reports of quality issues with staple software like Windows client seem to grow. Other groups have been given aggressive revenue goals or sales targets, forcing them to focus on increasing average revenue per user (ARPU) with new add-on licenses and reducing the avenues for customers to negotiate discounts or even directly do business with Microsoft. This includes the broad move of smaller Enterprise Agreement (EA) customers to the Microsoft Customer Agreement (MCA), and the discontinuation of existing discounts for online services provided across all of its EA tiers. 

And although there are undeniable use cases for AI, Microsoft’s relentless rebranding, repackaging, and false starts (Windows Recall, for example) have left customers confused. To make matters worse, heavily marketed new AI capabilities are often attached to Microsoft’s Frontier program which, similar to the older Windows Insider program, provides preview access to features that are unfinished and may never actually ship in a finished state. 

All of this makes doing business more complex for customers operating in an already uncertain environment, and can force unwanted change and uncertainty in places where organizations would prefer stability. But it pushes customers to where Microsoft wants them to be and helps move their AI efforts forward.  

What’s a Disgruntled Customer to Do? 

However, quitting Microsoft isn’t what most customers want either; nobody looking for stability wants that kind of complex change. The current situation emphasizes the importance of not taking the Microsoft relationship for granted or leaving it on “cruise control”, and not assuming that Microsoft will develop their products with your needs first and foremost in mind, especially if your company doesn’t fit the mold of AI-forward “Frontier Firms.” 

Customers need to take an active approach to managing their relationships with Microsoft around the services and software they use, focusing relentlessly on what benefits their own businesses even if it doesn’t align with Microsoft’s strategy. Research and analysis from independent sources (yes, like Directions on Microsoft) can be an invaluable tool. But it’s just as important to internalize that what your business needs and what Microsoft wants are probably not one and the same. 

Strategies exist that you can leverage, even if you’re not fully aligned with what Microsoft wants: 

Give a little, get a little. Microsoft’s sales force is getting high pressure to sell AI services. Use that to your advantage when negotiating your contracts. Maybe buy more M365 Copilot than you planned, in exchange for concessions on things you need like ESUs, exceptions for special use cases, or credits for services that your business already relies on. 

Avoid the FOMO. Yes, there are plenty of white papers and customer cases that Microsoft will tout, and maybe a magic quadrant or two. But there are lies, damned lies, and statistics. When Microsoft says everyone is using Copilot, do they mean the expensive M365 Copilot, or that every search typed into Edge is technically a “Copilot search?” Keep clear of false urgency, and focus on what best benefits your business needs. 

Let some of the dust settle. The landscape is still rapidly changing, even if M365 Copilot technically went GA in late 2023 (as Copilot for Microsoft 365!) Rebranding and repackaging is ongoing, agents have become the newest hot thing, many heavily promoted features are only previews, and industry alliances are still shifting. Act judiciously and focus on solutions with real benefit—the cost of adopting too early based on promises and previews can be higher than a missed opportunity. 

You can learn more about making the best of your relationship with Microsoft from the Directions on Microsoft Briefing podcast, or the Directions YouTube channel. 

What challenges is Microsoft’s AI monomania creating for your business, and how are you handling it? Email me at jgaynor@directionsonmicrosoft.com.  

Microsoft will be increasing pricing for its commercial Microsoft 365 and Office 365 suites beginning July 1, 2026.   

Company officials attributed the increase to the inclusion of AI capabilities such as Copilot Chat, along with security and management add-ons it is bringing to the Office and Microsoft 365 commercial suites. 

Microsoft 365 E3’s list price will increase from $36 USD per user per month to $39 USD per user pe month. Microsoft 365 E5 also will increase by $3 USD per user per month — from $57 USD to $60 USD. The Microsoft 365 F1 SKU is going from $2.25 USD to $3 USD per user per month; F3 is going from $8 USD to $10 USD per user per month. Officials said nonprofit pricing will be “adjusted in line with commercial pricing, as it is tied to commercial rates through a fixed percentage discount.”  Government Office suite pricing is going up, as well; in some cases phased over two years.

Update (Feb. 25, 2026): A Microsoft spokesperson had told us last month thathe company had no comment when we asked for a more complete updated price chart. But on Feb. 16, 2026, Microsoft posted a full list of the coming M365/O365 pricing and packaging changes.

In 2022, Microsoft increased prices of most of its Office 365/Microsoft 365 subscriptions, also citing the expansion of features as the reason. At that time, E3 went from $32 USD to $36 USD per user per month, but held the price of E5 steady at $57 USD per user per month  to try to increase usage of that particular version. In 2024, Microsoft also increased prices of its on-premises Office servers by 10%. 

More Built-In Features 

In a Dec. 4 blog post, Microsoft officials detailed some of the features the company recently added and/or planned to add to Microsoft 365 suites for no additional charge. Among those features: 

• M365 plans now include Copilot Chat Word, Excel, PowerPoint, Outlook, and OneNote. Microsoft will still contine to charge an additional $30 USD per user per month for a full M365 Copilot license (on top of the more expensive Office subscriptions). 

• Microsoft 365 E3 and Office 365 E3 are getting Defender for Office Plan 1’s enhanced email security features. 

• Microsoft 365 E3 and E5 are getting certain Intune capabilities, including Intune Remote Help, Intune Advanced Analytics and Intune Plan 2. E5 customers also get Intune Endpoint Privilege Management, Enterprise Application Management and Microsoft Cloud PKI. (These capabilities will continue to be available as separately priced add-ons.) 

• Microsoft 365 E5 customers are getting Security Copilot features. (Currently, only M365 E5 customers who purchased Security Copilot are getting these, but the plan is to make them available to all M365 E5 users “in the coming months.) 

Teams Pricing Tweaks 

In other related pricing news, Microsoft recently tweaked its availability and pricing for its Microsoft 365 suites with and without Teams.   

Starting Nov. 1, as the result of an agreement with the European Commission, new customers worldwide were able again to purchase M365 and Office Enterprise suites that include Teams. And per European Commission requirements, Microsoft increased the delta between the price of M365 and Office 365 suites with and without Teams. 

The new pricing:  

• M365 E3 (with Teams) is $36 USD per user per month 

• M365 E3 (no Teams) is $27.45 USD per user per month. 

• M365 E5 (no Teams) is $48.45 USD per user per month 

• M365 E5 (with Teams) is $57 USD per user per month 

• M365 F3 (no Teams) is $6.93 USD per user per month 

• M365 F3 (with Teams) is $8 USD per user per month 

(Note: These prices will all be increasing as of July 1, 2026, as noted above.) 

Microsoft also agreed to provide better interoperability and data portability as part of its agreement with the European Commission. The company said it will allow third-party communication and collaboration solution providers to access M365 and Teams the same way other software companies developing add-ins do.  

Microsoft also said it plans to make available application programming interfaces (APIs) that enable access to customer data from Entra ID, Exchange Online, SharePoint Online and OneDrive in a way that is “effectively equivalent” to what Teams accesses from these services. And it will provide APIs and tools to help customers export their information from Teams for use in other third-party solutions, if desired. Microsoft didn’t provide timing for any of these commitments. 

By Barry Briggs

Let the semantic-layer wars begin!

A little over a month ago, I made the less-than-prescient claim that “Microsoft has done little to nothing about a semantic layer.” But – in fairness to me – I also mentioned “rumors of a stealth mega-project underway at Microsoft to build some sort of ‘semantic layer’” that might compete with leaders in the field like Palantir, Celonis, AtScale, and others.

At Ignite two weeks ago, a succession of speakers elaborated upon a new brand, called “IQ,” that purportedly brings new intelligent capabilities to Fabric, Microsoft 365, and Azure AI Search, called, respectively, Fabric IQ, Work IQ, and Foundry IQ. (These appear to supersede an earlier capability called Context IQ, introduced in 2021.)

Let’s unpack these important announcements. First, we’ll discuss what they are and what they can do for you; then we’ll compare and contrast them with other offerings in the industry, notably Palantir, and what it means for them; and lastly, what’s likely to come next.

The Three IQs

As I mentioned, at Ignite Microsoft announced three new features under the IQ brand; each of these layer on top of existing products with the goal of further empowering both end users and the AI agents that Microsoft expects will become prevalent. Each, in different ways, builds a “knowledge graph” of the underlying data store – that is, links and relationships between disconnected pieces of data. Such a graph might, for example, connect a database of employees with a facilities database, enabling insights for how well office space is used, if teams are collocated, and so on.

In each of the three new capabilities Microsoft builds on existing layers.

Work IQ unsurprisingly leverages Microsoft 365’s Graph API – which already maintains many such relationships – and adds so-called “memory” features, learning and recalling users’ actions and habits. Thus a user might ask Microsoft 365 Copilot to “find the email from Susan a few weeks ago, put in the pricing she suggests into a contract like the one we did with Acme, customize the logos appropriately, and generate a Word document,” or “create a detailed financial model of the proposed acquisition of Contoso” in Excel.

In Fabric IQ, Microsoft again leverages an existing capability, in this case the semantic models created in support of Power BI applications, to create a knowledge graph of the department or enterprise. Alternatively, data professionals can create their own ontologies (maps or blueprints of the relationships between key subject areas); for example, product > part > supplier > contract > terms, using a new Fabric ontology tool (in preview).

Microsoft claims that Fabric IQ helps resolve core business concepts (“customer,” etc.) and illuminates relationships to other entities. (As we’ll see, this is harder than it sounds.)

Moreover, with two new agent types, Data Agents and Operations Agents (the former being in preview all year), enterprises can deploy sophisticated chatbots and even operational (IoT) agents that monitor devices in real time and predict and prevent failures, respectively. (Note that both of these agents are designed to work without Fabric IQ but have more capabilities with Fabric IQ.)

Finally, Foundry IQ, based on Azure AI Search (and built by that team), helps developers quickly index enterprise data sources and make them available to whatever agent might need them. Instead of a RAG application querying a single data source, Foundry IQ simplifies the ability for such apps to examine a range of sources.

The New New Application Stack

Back in April, I posited that Microsoft’s vision for the AI-centric application stack consisted of a UI layer (Copilot); logic (AI agents); data (Fabric, et.al.); and hardware (massive data centers).

We need to tweak the stack a bit: Microsoft has added a new “semantic layer” to its data sources, with the goal of enabling agents (and humans) to make increasingly better decisions. (I say “increasingly” because over time agents – in theory – learn and fine-tune their actions.)

What Could Possibly Go Wrong?

It’s a compelling vision, but it’s likely that the pesky realities of enterprise data will slow adoption. Few enterprises have a well-architected enterprise data model (EDM), as we discussed in our previous blog post. So – for example – what happens when the carefully designed definition for “customer” created in the ontology tool conflicts (say, in something as trivial as the number of address lines) with that in one or another system? Or when a salesperson insists on one definition and the finance analyst, another?

Creating an ontology, in Fabric or elsewhere, is a major effort and requires on-going maintenance, so there’s no getting out of the hard work. For example, the Fabric ontology tool can discover Azure hosted data and build the foundation of an ontology on what it sees, but bringing in external data sources like an Oracle cloud database takes manual effort.

Even with auto-discovery, the people you charge with building and maintaining the ontology will face a hill of challenges, determining how one data source relates to another, what column names to use, how to resolve ownership of data that appears in two sources, creating additional dimensions (data from data) to address a business demand, and configuring triggers that allow data in one source to act on data in another source. And what happens when one of the external data sources inevitably changes? And, they have to do all this while following the security requirements and data sovereignty of each data source.

Still, the appearance of semantic-layer technologies across Microsoft’s portfolio can enable some powerful usage scenarios – yet another reason, perhaps, for enterprises to embark on data hygiene projects. (I recommend, based on experience, to start at the department level before taking on something as massive as a full EDM: quick wins to show value and focusing on where the data is created can accelerate adoption and convince skeptics.)

The Semantic Wars: Who Do You Back?

Since the announcement of the “IQs,” lots of virtual ink has been spilled comparing them to established vendors, most notably Palantir. Here’s a quick summary of the situation as it stands today:

• Work IQ, Fabric IQ, and Foundry IQ are new (with many of the features still in preview) and not yet widely adopted; however, for Microsoft-centric enterprises they’re likely the right choice, since they build upon and leverage existing features in the underlying platforms and services.
  
• The IQs are three disconnected services which – frankly – highlight Microsoft’s organization chart and will add complexity to creating agents that must span all enterprise data.
  
• Palantir’s Foundry (yes, the same name) is more mature and widely adopted, particularly in operational IoT scenarios. Its strategic focus is to help enterprises build a “digital twin” of their operations.
  
• Palantir’s metamodel consisting of entities and relationships (semantic layer); actions and functions (kinetic layer); and security and permissions (dynamic layer) is well thought out and cohesive. But Palantir relies on a hands-on consulting engagement with their own “forward-deployed engineers” or Accenture, Deloitte, and/or others to create the specific instances of the metamodel, which can be expensive and time-consuming.

What’s Next?

There’s undoubtedly more to come in the world of IQ. I’d hardly be surprised to learn of a “Purview IQ” which could leverage Purview’s Unified Data Catalog to reach into non-Microsoft data like Snowflake or Databricks, or even data in other clouds.

Similarly, I’d also expect IQs on each of the Dynamics 365 applications – Finance, Supply Chain Management, HR, and so on, with a Copilot UI on each of them. (Let’s just hope the “IQ” brand doesn’t get quite a dilute as “Copilot.”)

Are Systems of Intelligence Here?

Whew!

My personal opinion is that of all the announcements at Ignite, the IQs will become the most important and consequential. Perhaps indeed we are at the dawn of what Geoffrey Moore and George Gilbert called (all the way back in 2015) “systems of intelligence” that can “[mine] engagements in real-time to anticipate, influence, and optimize customer experiences.  In their most advanced form, these systems learn both from the user’s interactions as well as external observations such as an extended network of relationships.”

Do I need an IQ layer? Drop me a line at bbriggs@directionsonmicrosoft.com.

Microsoft’s annual Ignite conference for IT professionals, developers, and partners, is kicking off this week in San Francisco. CEO Satya Nadella is not the lead keynoter this year; instead, the newly minted CEO of Microsoft’s Commercial business, Judson Althoff, is doing the honors. (Nadella is busy extolling the virtues of Microsoft’s sprawling network of “fungible fleets” in its new AI superfactories.) 

To no one’s surprise, this year’s Ignite will focus on all things AI-related — and especially agent-related.  

Microsoft announced ahead of the start of Ignite on Nov. 18 that it is adding Anthropic’s models to Microsoft Foundry (the rebranded Azure Foundry), giving the company bragging rights to being “the only cloud providing access to both Claude and GPT frontier models to customers on one platform.” Microsoft execs said they will provide access to Claude across Microsoft’s Copilot family, including GitHub Copilot, Microsoft 365 Copilot, and Copilot Studio. Microsoft committed to invest up to $5 billion in Anthropic; Anthropic committed to purchase $30 billion in Azure compute capacity from Microsoft.

Directions‘ Top Ten Ignite Picks 

Here are our picks for the top 10 announcements that could impact enterprise customers and partners moving forward into 2026. In no particular order: 

1. The Agent 365 control plane: Microsoft inadvertently published some information about Agent 365 a week ago. As anticipated, this “control plane” is all about making autonomous agents more centrally manageable and secure by treating them more like user accounts. Agent 365 includes the Agent SDK, which is required to build Agent 365-aware agents and publish them to the Agent Store. These Agent 365-aware agents will integrate with Microsoft security controls, policies, compliance capabilities, and Office apps without needing custom integrations, Microsoft says. The Entra Agent ID piece is available in the Microsoft 365 admin center for those in the “Frontier” early-access program. We’ve asked Microsoft about any additional licensing and pricing required for Agent 365, but no word back so far. 

2. Work IQ links agents to work data: Microsoft has been working on adding AI smarts to its Office apps since 2021, if not before, under the “Context IQ” program. Work IQ looks to be either a rebrand or an evolution of Microsoft’s Context IQ technology.  Like Context IQ, Work IQ is a layer that connects to organizational and personal data such as files, emails, meetings, etc. and can “build memory” based on users’ preferences, habits, and workflows. As Context IQ is already part of the Microsoft 365 stack, Work IQ already is built in, as well. But now developers can access Work IQ to build agents via APIs. 

3. Even more IQ madness — Fabric IQ and Foundry IQ: There are other “IQ” branded layers in the Microsoft AI stack that are meant to help agents understand what users are doing, where to find information, and what the business data means. Fabric IQ, now in preview, extends the unified semantic layer already provided in Power BI across 20 million models to business operations, Microsoft says. Foundry IQ, also in preview, is built on top of Azure AI Search and is meant to be the “next generation of retrieval-augmented generation (RAG),” according to the company, providing a way to connect agents with data from a single knowledge base.  Microsoft is rebranding its family of Azure AI services, Logic App Connectors, and MCP servers as “Foundry Tools.”

To understand why these IQ pieces matters at a higher level, Directions analyst Barry Briggs recently blogged about why a semantic layer is so crucial to making AI useful. 

4. Microsoft 365 E5 users get Security Copilot: Security Copilot will be included for all Microsoft 365 E5 customers. Eligible E5 customers will get 400 Security Compute Units (SCUs) per month for every 1,000 licensed users (US$1,600 in value), up to 10,000 SCUs per month (US$40,000 in value, with 25,000 E5-licensed users). If you need more, it’ll be $6 per SCU on a Pay-As-You-Go basis.) Microsoft will begin offering this bundle starting today for existing E5 customers who already have Security Copilot and for all other E5 customers “in the upcoming months.” Security Copilot will continue to be available for customers without E5 separately.  

5. “Agent Factory” plan bundles Microsoft Foundry, Copilot Studio: As Microsoft has continued to add more “pro” capabilities to its Copilot Studio agent-building platform, it’s also been building out agent creation and management features in Microsoft Foundry. It is introducing a new “metered” Agent Factory plan that will allow developers to build and manage agents in Foundry and/or Copilot Studio and deploy their agents anywhere, including Microsoft 365 Copilot. We’ve asked for more details on how this new plan will be licensed and priced, but no word back so far. Update (Nov. 18): A Microsoft spokesperson said the company had “nothing to share” on how the new metered plan will be priced/licensed or if it will be required for developers who want to tie into Work/Fabric/Foundry IQ.

Update (Nov. 18): Aha! It looks like Microsoft’s recently announced Agent Pre-Purchase Plan (P3) is key to how Agent Factory service will be priced. P3 will let customers access both Copilot Studio and Foundry services through one pool of funds. Customers will be able to buy Agent (ACUs) up front in a one-year metered plan.

6. A cheaper Microsoft 365 Copilot plan for SMBs: Microsoft quietly has been offering some hefty volume discounts to its largest customers interested in Microsoft 365 Copilot, priced at $30 per user per month. Starting in Dec. 2025, it also will offer a monthly $21 per user option for customers with fewer than 300 users. The new plan is called Microsoft 365 Copilot Business. 

7. New Dedicated Word, Excel, PowerPoint agents: Microsoft 365 Copilot will feature dedicated Word, Excel and PowerPoint Agents — first available through the Frontier test program — that can create Office content directly from Microsoft 365 Copilot Chat. If you’re confused about the difference between the existing Office Agent, these new single-app agents and the existing “Agent Modes” in various Office apps, join the club. We’ve asked Microsoft to explain what’s what. No word back so far. Update (Nov. 18): Microsoft’s blog post about the coming Office agents sheds (very little) light on what these things are. But it seems the new Anthropic-based agents are going to be available with M365 Copilot Chat, rather than in the individual Office apps, as the “Agent Modes” are. Microsoft believes customers will start in Chat with the agents and then move to the Modes to complete tasks.

8. Windows 365 to become an agent destination: Windows 365 for Agents will allow developers to build and deploy agents that can do things like open apps, process data and automate tasks on Cloud PCs. Developers can access Copilot Studio in Windows or the Researcher agent in Linux in their secure Cloud PCs. Developers interested in trying out Windows 365 for Agents can join a waiting list. 

9. SQL Server 2025 is GA: The latest version of SQL Server, which Microsoft calls “the AI-ready enterprise database,” is now generally available. This release provides advanced semantic search and improved natural-language access. It also features updated identity and encryption integration, GitHub Copilot integration, and support for database mirroring in Fabric. Microsoft says the 2025 release is the most significant SQL Server release in a decade for developers. In other database news, Microsoft says Azure DocumentDB (at one point called “Azure Cosmos DB:) is now GA and and Azure HorizonDB, a new PostgreSQL cloud database service for building or modernizing mission-critical apps, is now in private preview.

10. Windows as an “agentic OS”: Because no Microsoft product can advance without an AI story, Windows 11 is in the process of becoming an “agentic OS” — much to the opposition of many long-time Windows loyalists. Microsoft is adding its own agents and an “Ask Copilot” link to the Windows taskbar to try to increase Copilot usage. It’s also building in agent connectors (like a File Explorer and Windows Settings connector); and testing in private preview an Agent workspace where agents can interact with software in an isolated, policy-controlled, auditable environment. Model Context Protocol (MCP) support in Windows 11, in preview, will let agents to access those connectors to complete tasks for users. 

Ahead of Microsoft’s Ignite conference next week, rumors are swirling about Microsoft launching an “Agent 365” platform. If the tips pan out, Agent 365 could become a way for Microsoft to unify how the increasing number of autonomous agents will be managed, licensed and priced across Microsoft 365, Dynamics 365 and the Power Platform, going forward. 

Microsoft seemingly inadvertently published a screen shot featuring the “Agent 365” branding earlier this month as part of a Microsoft 365 Admin Center post (which the company has subsequently unpublished). Microsoft also removed the connected roadmap entry (518220) about agentic users from its Microsoft 365 Roadmap.  

The screen shot was featured in MC1183300, which Microsoft initially published on Nov. 6 (and a copy of which is here). That message described “agentic users” as “a new class of AI powered digital entities designed to function as autonomous, enterprise-grade virtual colleagues.” Agentic users, unlike bots, will be provisioned as “full-fledged user objects,” with identities in the organization’s directory (via Entra ID or Azure AD), email addresses, Teams accounts and presence in the org chart. 

Agent templates, which are preconfigured digital worker profiles, will be available in Microsoft Teams and the Microsoft 365 Agent Store, according to the message. Targeted rollout of these agentic users is supposed to begin in mid-November on the desktop platform for Teams and Microsoft 365 Copilot, the message says. And admins will be the ones assigning the required “A365” license at the time of approval, with “no additional Microsoft 365 or Teams license required.” 

In early Oct. 2025, Microsoft announced the Microsoft Agent Framework, which is an open-source software development kit and runtime for building and managing multi-agent systems. And this past spring, Microsoft took the wraps off another piece of the supporting puzzle: Microsoft Entra Agent ID, which is designed to help manage agents’ identities, lifecycles and permissions across Microsoft Copilot Studio, Azure AI Foundry, and later, Security Copilot, Microsoft 365 Copilot and third-party solutions. 

Pricing and Licensing Details Not Yet Clear 

While Microsoft has priced Microsoft 365 Copilot at a flat $30 per user per month, it hasn’t delivered a clear, consistent approach to licensing and pricing agents.  

It has folded some of its agents, like Researcher and Analyst, into Microsoft 365 Copilot for no additional charge. But in the case of SharePoint Agents and Copilot Studio, which customers can use to build agents, agents are priced based on a number of factors. Agents are charged via meters that count the consumption of a Microsoft-defined unit called Copilot Credits (formerly “messages”). Copilot Credits are billed on a pay-as-you-go (PAYG) basis or by purchasing capacity packs up front for a discount. In Oct. 2025 (called Pre-Purchase Plan or sometimes P3) allows upfront purchasing of Copilot Credits on an annual basis with discounts that increase with the number of Copilot Credits purchased. 

It does seem as though the new Agent 365 licensing model could apply to both Microsoft-built and custom agents. And under the new model, every agent would need a license, in the way that every user needs one now. 

Rich Gibbons, of Cloudy with a Chance of Licensing fame, has wondered whether Microsoft is planning to price and license its agents by role, meaning there could be an Agent 365 for HR, for Sales, etc. Or maybe Microsoft is looking to enable agent pricing and licensing based on more granular job functions with certain industry domains, like booking patient appointments in the healthcare field, as Dynamics 365 Consultant Muhammad Hammad Wali suggested in a post on LinkedIn. 

Hopefully Ignite will bring answers to at least some of our Agent 365 questions. 

Microsoft’s continued huge and growing capital expenditure spending was the big story this week. For customers, the key take-away from Microsoft’s Q1 FY’26 earnings is Microsoft is still in a supply-constrained state for cloud compute, with its executives saying they don’t see the situation changing until sometime after June 2026. The extent to which this will affect current and future Azure customers isn’t clear. 

Microsoft’s Q1 revenue was $77.7 billion, up 18 percent, and net income at $27.7 billion. The Intelligent Cloud business, which includes Azure, brought in $30.9 billion, up 28 percent. Microsoft rarely discusses Azure revenue figures, though last quarter execs did say Azure and related cloud services surpassed $75 billion in fiscal 2025. Microsoft also no longer publicly discloses the point contribution of AI to its Azure numbers. 

Last quarter, Chief Financial Officer Amy Hood said Microsoft expected supply constraints to ease in fiscal 2026. But that’s no longer the case. 

“We know we’re behind. We do need to spend,” said Hood. 

In fiscal 2026, which began July 1, Microsoft plans to increase AI capacity by 80 percent and nearly double its datacenter footprint over the next two years, CEO Satya Nadella said. Microsoft currently has more than 400 datacenters; its newest ones tend to be larger, square-footage-wise. 

Hood said Microsoft will increase its spending on GPUs, CPUs, and datacenter leases in fiscal 2026. In Q1, Microsoft’s capital expenditures were $34.9 billion, higher than the $30 billion officials were predicting last quarter.  

Hood said roughly half of Q1 capex spending went toward CPUs and GPUs that support increasing Azure platform demand, first-party apps and AI solutions and R&D by its product teams, plus replacement for end-of-life server and networking equipment. The other half was for “long-lived” assets that will support the platform for the next 15 years and beyond, including finance leases for large datacenter sites. 

Still No M365 Copilot Sales Disclosures

Commercial bookings were up a whopping 112 percent this quarter — “significantly ahead of expectations,” Hood said. A substantial part of this  is from Azure commitments from OpenAI, she said, as well as continued growth in the number of 100-million-dollar-plus contracts for both Azure and Microsoft 365. (It’s worth noting these numbers do not include the $250 billion Azure commitments from OpenAI that the two companies announced earlier this week, as part of the details of their new partnership agreement.) 

Unsurprisingly, Microsoft still did not disclose sales numbers for Microsoft 365 Copilot during its Oct. 29 earnings call. Officials said the company’s “family of Copilots and agents” surpassed 150 million monthly active users across Microsoft 365 Copilot, GitHub Copilot, Security Copilot and other “high-value domains.” Last quarter, officials said that number was 100 million. 

For what it’s worth, officials also claim that Microsoft now has 900 million monthly active users of our AI features across its products. 

Microsoft-OpenAI: A Bit More Peace of Mind 

Speaking of the updated Microsoft-OpenAI partnership deal, there seems to be little new that will affect Microsoft customers directly, beyond some possible peace of mind that Microsoft will continue to have access to OpenAI IP for several more years. 

Microsoft is getting a 27 percent equity stake in OpenAI’s new for-profit entity. Microsoft no longer has a “right of first refusal” on new OpenAI cloud workloads, but it will retain IP rights to OpenAI models and products through 2032. OpenAI can jointly develop some products with other companies, but API products are exclusive to Azure. And once OpenAI declares it has achieved the elusive Artificial General Intelligence (AGI) mark, which affects Microsoft’s access to its technology, a panel of unnamed experts will have to verify that claim. 

“This punts into the long grass the idea that Microsoft might lose its exclusive access to OpenAI’s fundamental models. Specifically, OpenAI can no longer just declare it has reached AGI and use that to bring on other API providers,” said Directions on Microsoft analyst Greg DeMichillie. 

“If you are using Azure OpenAI but were afraid that somehow OpenAI would pull the rug out from under you and you’d be left stranded, that doesn’t seem like a concern anymore,” DeMichillie added. 

Microsoft disclosed that its investment in OpenAI resulted in a $3.1 billion ding to net income in the quarter. By The Register’s calculations, that could mean that OpenAI itself lost $11.5 billion for the quarter. 

This blog post is a little different. I’m not going to tell you about the pros and cons of a Microsoft technology – rather, I’m going to talk about a missing Microsoft piece of the puzzle.

Maybe it’s even the Redmond giant’s Achilles’ heel.

As AI transforms practically everything about how we work, it’s become clear that a well-managed and well-governed data estate is critical for its success in the enterprise (see my video). But even with data that’s appropriately secured, and whose regulatory compliance is constantly monitored with Purview and other tools, the question remains: how can AI put all this data work?

Sisyphus and the Enterprise Data Model

We in IT have struggled for decades with so-called “islands of data,” that is, incompatible clumps of information squirreled away in ERP, CRM, Support, Legal, IoT, e-Commerce, and other systems. One application’s “customer” is another’s “contact” or yet another’s “client,” or “lead,” or “account.” Some keep two address lines for the individual, some four; some have a 5-digit ZIP code (in the US), some nine; and keeping them all in sync is, in a word, a nightmare that can cost enterprises millions.

To simplify this, information architects and data managers have struggled for years to build an “enterprise data model:” common definitions of key subject areas (“customer,” “product,” etc.). With such common definitions – a blueprint, if you will – all departments and applications can have a shared understanding of the subject area, can better achieve compliance goals, can improve communications, and can streamline application integration efforts.

And there have been many attempts to create industry-standard models, and some have been useful. But there’s a key missing point here, which is that data models encapsulate a company’s competitive differentiation. (Think about it. It’s true.) Standardize the data model, eliminate your differentiation.

Because of that, data models that affect competitive differentiation – how a company thinks of customers, products, pricing, discounts, and so on – necessarily change rapidly, much faster than an EDM could track.

Yet…the dream remains…wouldn’t it be nice if there were one single data model for each subject to which all applications aligned?

Well, yes. But it’s been an uphill battle (thus Sisyphus). Why? Creating and deploying an EDM is excruciatingly expensive, painful, and difficult, both in terms of the human cost of negotiating common models as well as the technical costs. While potentially streamlining integration (no more “T” in ETL!), simplifying process modeling, and other cost-saving benefits, the cost and risk of updating applications schemas is so high that it’s never been seen as a worthy effort.

Until now, maybe?

Maybe, Just Maybe, the Enterprise Data Model is Possible

But imagine – just for a moment – if you could describe, even at a very high level, your core business entities. (A quick note on terminology. When we (and vendors) say “subject area,” “business object,” and “business entity” we mean pretty much the same thing: the nouns that run your business, as in a customer buys a product. See how hard it is to agree on terminology?)

Say business folks and IT staff can come together and decide, “here’s the authoritative definition of a retail customer” for our company.

Then (keep imagining) you push a button and automagically, using automated connectors, software reaches out to applications and maps your definition to real data in the real world. And maybe it (perhaps with human help) creates relationships (a retail customer “is a” kind of customer, “has an” account manager, and so on).

Companies like Celonis, Palantir and AtScale propose to do just this. On top of the oceans of data collected in Snowflake, Databricks, or Fabric, then, these models form the foundation of what my friends George Gilbert and Peter O’Kelly call a “semantic layer,” imposing order on chaos.

The Enterprise Digital Twin: Stir in AI, and Magic Happens

The vision for both companies: to create an AI-powered “digital twin” of the entire enterprise. Think of it as the organization’s automated pan-intelligent COO: you can issue it commands, you can ask questions of it, and expect it to know everything that’s going on, everywhere in your business.

Greatly Simplified View of Enterprise Ontology

Celonis uses sophisticated process mining to discover both how data is used in processes, and the often myriad variations of processes hidden in the data. By shining a light on enterprise processes — remember, a company is nothing less, nothing more than a collection of processes – you can optimize them. Where are the bottlenecks? Why, for example, when a customer orders a widget does it sometimes take weeks to fulfill?

Whereas Palantir focuses on prediction, simulation, and coordination based on holistically linked real (and real-time) data. So if CEO wants to know “how will raising the price of a widget by 5% in Japan affect revenue and profitability?” – in theory with such a digital twin those sorts of questions could be answered.

Color Me Skeptical

Now at this point I have to inject some personal experience and opinion. I’ve been involved with efforts to build an EDM in large enterprises; it’s unspeakably hard, for all the reasons noted above. It’s very labor-intensive and detail-oriented: at one company I recall seeing a model for “customer” printed on a plotter that stretched out across an entire conference room table (in 8-point font!). Fundamentally: useless.

I’m willing to keep an open mind – but for any such endeavor like this I would caution all: show ROI quickly. Do pilots. Look for quick wins. Focus on department processes, not the full enterprise to start. Demonstrate the value as quickly as possible!

Where is Microsoft?

Now I’m sure you’re saying: Barry, this is the Directions on Microsoft blog, and you’ve barely mentioned Microsoft!

True! Microsoft, of course, has made massive investments in data management, most recently with Fabric, its answer to Snowflake and Databricks (and others). And let’s be clear, what’s different about today is the presence of these massive data lakes, lakehouses, and so on, that consolidate data in ways we were never able to before.

But, to date, Microsoft has done little to nothing about a semantic model. Purview does have a Business Glossary of common business terms – but it’s not linked to real data in any way. Mary Jo reminds me that they bought a process mining company called Minit back in 2022 – it, however, seems to have been folded into Power Automate. And (just to confuse matters) there’s Flow Builder, which industry watcher Jukka Niiranen speculates could replace Power Automate. But Power Automate – like the Power Platform generally – isn’t targeted at enterprise-scale, enterprise-class scenarios. And it doesn’t pretend to create a “semantic layer.”

And yet this semantic layer may be the thing that truly unlocks the power of AI in the enterprise.

I’ve heard rumors of a stealth mega-project underway at Microsoft to build some sort of “semantic layer” that competes with Palantir and Celonis. Is it real? Is it good? Is it Minit reborn? (If you know, drop me a line). We’ll see.

For now, it seems like a gap. And it may be a big one.

Do I have gaps in my thinking? Drop me a line at bbriggs@directionsonmicrosoft.com. And if you haven’t read The Laws of Business Process, well, you really must.

Just a couple of weeks ahead of the end of support for Windows 10, Microsoft has started rolling out its most recent Windows release: Windows 11 25H2. This is a fairly minor update with only a handful of features of potential interest to enterprises.

At the moment, Windows 11 25H2 is available to “seekers,” meaning it requires users to either search for updates or to download and install it. The 25H2 release also is available via Windows Autopatch; the Microsoft 365 Admin Center; Microsoft’s Software Download Service; and Visual Studio Subscriptions. Version 25H2 will become available via Windows Server Update Services (WSUS) on Oct. 14, 2025, with the October security update. After that, over the coming months as it usually does, Microsoft will broaden the pool of devices deemed as “eligible” for the release and ultimately deem it ready for enterprises to install.

Microsoft is delivering Windows 11 25H2 via an enablement package to those already running Windows 11 24H2. This means 25H2 is based on the same code base and servicing branch as 24H2, and the enablement package (eKB) will make moving to the latest release quicker and hopefully less problematic than a full OS refresh. Those on earlier versions of Windows will have to do a full system upgrade to get 25H2, however.

(Note: Microsoft used this same enablement-package strategy when introducing Windows 11 23H2, but not when it rolled out 24H2. The official reason for the pause with 24H2 was Microsoft changed too much of the Windows “foundational” code base, mainly by adding a lot of AI-focused functionality, to deliver the new version via an enablement package.)

A Handful of Brand-New Features

Microsoft has been rolling out new enterprise-focused features for Windows 11 24H2 on a regular cadence (which it calls “continuous innovation”). The company already released features like Quick Machine Recovery and hot patching for Windows Enterprise clients as part of these earlier feature updates, and these features are baked into 25H2.

On the very short list of what’s brand-new as of Windows11 25H2:

• Ability to remove certain inbox apps using Intune or Group Policy (for Windows 11 Enterprise and Education only), including Xbox gaming apps, Paint, consumer Outlook and more
• Wi-Fi 7 Enterprise profile supported
• Removal of PowerShell 2.0 and Windows Management Instrumentation command-line (WMIC) – which potentially could break some custom admin scripts
• “Advancements in build and runtime vulnerability detection, coupled with AI assisted secure coding” (with no additional explanation from Microsoft as to exactly what these features entail)

Windows 11 25H2 also includes several AI features that only work on Copilot+ PCs with dedicated neural processing units (NPUs), including Windows Recall, Click To Do, and enhanced Windows Search.

As detailed on its Windows IT Pro blog, Microsoft also has released the Windows 11 25H2 Security Baseline, Administrative templates (ADMX), and an Evaluation edition of Windows 11 25H2.

 Windows 11 Enterprise and Education customers get 36 months of support for the 25H2 release, starting September 30. (Home and Pro users get 24 months of support, as usual.)

He’s not a household name like Bill Gates, Steve Ballmer or Satya Nadella. But Judson Althoff — Microsoft’s latest exec to get a “CEO” title — now runs the entirety of Microsoft’s Commercial business. And that business comprises most of the company (about $220 billion of the company’s $282 billion in revenues in fiscal 2025). 

Althoff has been at Microsoft for almost 13 years. Before that, he was with Oracle, heading up worldwide alliances, channels and embedded sales for 11 years. 

In 2021, Microsoft already had begun moving Althoff up the org chart. At that time giving then-Executive Vice President Althoff the reins of Microsoft unified Global Sales and Marketing with its Worldwide Commercial Business. With that move, Microsoft was doing what a number of its competitors were: Flattening the hierarchy (a bit) and consolidating leadership. 

Different company watchers have different takes on the significance of Althoff’s latest promotion, announced on Oct. 1. 

Some are saying they see the move as a succession set-up play. As far as I know, there hasn’t been much, if any, public chatter about Nadella being ready to cede the CEO crown after 11 years in that role. But by giving Althoff the expanded role of leading commercial sales, marketing, operations, and engineering globally, Microsoft might be giving him a test-run as the next CEO of the company, some speculate.  

In his memo about Althoff’s promotion, Nadella said the reorg will “allow our engineering leaders and me” to focus on technical work “across our datacenter buildout, systems architecture, AI science, and product innovation.” Maybe this transition is the precursor to Nadella making a long, slow exit, similar to the way Gates became Chief Software Architect, relinquishing the CEO title to Ballmer back in 2000 (and only finally giving up his day-to-day duties in 2007)? Or maybe Nadella and Althoff will be co-CEOs at some point, with Nadella remaining the board chair? 

Nadella in (Non-Founder) Founder Mode? 

GeekWire had an interesting take, postulating that Nadella will now be freed up to take on more of a “Founder Mode” position, meaning he will be able to have more of a hands-on approach to areas (AI, AI, and more AI) in which he’s most interested.  

“Nadella’s always been more engineering-focused. It’s one of the things people talked about when he succeeded Ballmer,” said Directions on Microsoft analyst Jim Gaynor. “Meanwhile, Althoff has success in sales and commercial business. It really can just be a matter of placing Althoff where he’s strong so Nadella can focus on his own strengths and passions where they benefit Microsoft.”   

Althoff is not the only exec at Microsoft besides Nadella to have been granted a CEO title. Phil Spencer, CEO of Gaming, Ryan Roslansky, CEO of LinkedIn, and Mustafa Suleyman, CEO of Microsoft AI, are part of that club, as well.  

Althoff’s promotion comes on the heels of his two-month sabbatical following the latest round of Microsoft layoffs of an estimated 9,000 people, many of them in Althoff’s sales organization. 

By Barry Briggs

Compliance covers an ever-increasing, absolutely mandatory spectrum of activities, processes, and technologies with which your enterprise must concern itself.

The Compliance Imperative

Why compliance? Well, to start with, the penalties for noncompliance today can be severe. Just to note a few:

• Violations of the General Data Protection Act (GDPR) can mount as high as 4% of your company’s annual revenue. (Yes, GDPR is a European regulation, but if you do business in Europe, then you’re affected. And many countries and US states have followed suit with similar regulations.) Just ask Meta, hit with a €1.2b fine in 2023.
• Violations of the CAN-SPAM Act, which regulates unwanted email (all those opt-out and unsubscribe messages you see at the bottom of marketing emails? – that’s in response to CAN-SPAM) can reach over $50,000 per email.
• Violations of federal banking laws, such as anti-money laundering (AML) monitoring, have resulted in fines in the billions of dollars.

Fact: Compliance Has to Compete

Today’s CIO has to juggle numerous competing priorities, from adopting disruptive technologies like AI, to cybersecurity, maintenance (usually a huge and necessary part of any IT budget), technical debt, incremental feature improvements (like those in Power BI) and so on. Increasingly, however, regulatory compliance plays a role in every single IT expenditure and priority.

As AI becomes a key and pervasive mandate through nearly every enterprise, it opens up huge new risk areas for the compliance professional.

AI Changes Everything: Or Does It?

Recently I spoke with a senior compliance manager in one of the world’s top consulting organizations. When I asked how AI changes his job, his answer surprised me. “At one level,” he said, “AI changes nothing.”

He explained that the basic principles of compliance haven’t changed. Organizations must define a company-wide strategy for compliance by understanding which regulations and standards apply, who is accountable, and the tools and processes by which compliance will be assured; then continuously monitor activities to prevent intentional or inadvertent violations; and react to both internal activities (such as copying a sensitive file to a USB stick) and external requests (such as eDiscovery or a GDPR Subject Rights Request).

He went on to say, however, that AI increases the surface area that the compliance professional must manage and oversee. It’s certainly possible for individuals to use AI for (for example):

• Asking Copilot investment advice – and possibly (intentionally or not), pursuing insider trading;
• Finding inadequately protected sensitive data and exfiltrating it;
• Instigating discriminatory hiring practices with biased AI screening;
• Infringing on copyrights and/or plagiarizing;
• Violating patient, user, or employee privacy;

And so on. Many of these can happen even with the best of intentions so it’s key to develop comprehensive training and processes and use technologies to prevent them.

That’s where Microsoft Purview comes in.

Is Purview a Compliance Panacea?

Purview is Microsoft’s umbrella brand for a variety of risk and compliance related services, grouped and shown below. But not even all these varied capabilities suffice for the compliance professional.

There are two reasons Purview alone isn’t enough:

• You’ll need more than just Microsoft Purview services for compliance: you’ll need Entra (to manage user identities and access); Defender, for security incident management; Priva, for privacy management, and Azure services for roles and permissions. You might also need to access specific services via their own portal, such as Fabric, for application-specific, fine-grained controls.
• No doubt there are industry-specific regulations which apply to your organization for which you’ll need third-party tools. For example, banks and financial services institutions are required to maintain certain cash reserves; healthcare firms must have and maintain Business Associate Agreements (BAAs) with vendors; and so on.

Purview’s Expanded Purview Means You Need E5

All that said, Microsoft has made tremendous investments in Purview over the past few years; its capabilities have grown significantly and there’s lots that both use and address AI in the enterprise. But – there’s an essential truth here – effective compliance at the enterprise level pretty much requires a Microsoft 365 E5 license. Microsoft’s advances and innovations come at a cost – and, sorry to say, you’ll probably need them.

Here are a few examples:

The Compliance Manager runs “assessments” against over 300 international regulations and provides a score and suggested actions for improvement. But E3 only assesses against a generic, Microsoft-supplied amalgam of various regulations – a sort of least common denominator. If you want assessments against specific regulations – like PCI-DSS or the EU’s AI Act, or any others, you’ll need E5.

Information Protection, aka data classification, lets you assign labels to your content – a foundational part of data governance. But if you need to do it at scale – and who doesn’t – across SharePoint libraries, or by using AI to classify: you’ll need E5. Similar restrictions apply to classification’s cousin, Data Lifecycle Management, that provides data retention capabilities. (Pro tip: remember classification and retention are separate services.)

Similarly, Purview’s eDiscovery service offers features like reconstructing Teams conversations and accessing Microsoft 365 Copilot interactions – only in E5. Core features are offered in E3: but make sure your legal and compliance teams think they’re sufficient. They’re probably not.

Insider Risk Management – only available in E5 – helps organizations detect risky behavior, be it malicious or unintentional, such as IP theft or regulatory violations.

However, one piece of good news is that Purview’s new Data Security Posture Management (DSPM) for AI is available in lower tiers of Purview.

Think of DSPM for AI as a dashboard consolidating information from across the organization. You can use it to monitor AI usage, including with third party sites like ChatGPT and Claude; to see where sensitive data was accessed by AI tools; and to see where “risky” chats might be happening. It’s useful – and commendable – that this is available to non-E5 organizations.

Gotchas and Other Things to Know

While, as we’ve said, the principles of compliance haven’t changed, AI expands the surface area that compliance managers and tools have to monitor. There are a few things you should remember:

• Purview’s services are impressive. But they’re next to useless unless your employees have been trained, both in the importance of compliance and in the necessity of using the tools.
• AI eats data everywhere it finds it. It’s vitally important to protect your data estate as I described in my data governance video.
• To get the most from Purview, you’ll need E5. But remember a key Directions on Microsoft maxim: everybody who benefits from an E5 license must have one.
• For many of its services, Purview scans Microsoft 365 data and makes decisions – if it violates regulations, suggests risky behavior, and so on. Beware, however, of false positives.

Finally, as I’ve noted, remember that Purview is only one of likely many tools you’ll use for compliance. (In a recent webinar, I polled the audience and found over half used five or more compliance technologies.)

Disagree? Think I should comply with shorter blogs? Drop me a line at bbriggs@directionsonmicrosoft.com!

Microsoft is adding Anthropic’s models to Microsoft 365 Copilot, alongside the existing OpenAI models that are already at the heart of it. Microsoft is kicking off this process by adding Anthropic’s Claude Sonnet 4 and Claude Opus 4.1 as model choices for those using the M365 Copilot Researcher Agent and for those building agents in Microsoft Copilot Studio. Claude will be coming to more elements of M365 Copilot in the future, Microsoft says. 

Some reports claim Microsoft is doing this because the company, via its own testing, has found Claude to generate better results than OpenAI’s GPT. Other reports claim Microsoft is expanding the available models as a way to protect itself against being too dependent on OpenAI. Microsoft’s official reason for the move: It “advances our commitment to bring the best AI innovation from across the industry to Microsoft 365 Copilot.” 

Regardless of the reason, commercial customers interested in using Claude alongside and/or instead of GPT-based models in M365 Copilot and Copilot Studio definitely need to read the fine print. Because Claude is hosted on AWS, not Azure, by using Claude customers forego a lot of the protections they get when using OpenAI’s GPT and GPT-based solutions. 

According to an article on connecting to AI models on Microsoft Learn: 

“When your organization chooses to use an Anthropic model, your organization is choosing to share your data with Anthropic to power the features. This data is processed outside all Microsoft‑managed environments and audit controls, therefore Microsoft’s customer agreements, including the Product Terms and Data Processing Addendum do not apply. In addition, Microsoft’s data‑residency commitments, audit and compliance requirements, service level agreements, and Customer Copyright Commitment do not apply to your use of Anthropic services. Instead, use of Anthropic’s services is governed by Anthropic’s Commercial Terms of Service and Anthropic’s Data Processing Addendum.” 

The Agent Rollout Parade Continues 

It’s not just on the model front that Microsoft has been adding more customer choice. Last week, Microsoft provided an update on some of the many agents it first introduced in Nov. 2024 at its Ignite conference (or, in some cases, even before last Nov.).  

Microsoft is building out a fleet of agents designed to be embedded in Teams, SharePoint and Viva Engage, among its other products. The company is trying to improve the value proposition of its Microsoft 365 AI tools by enabling them to work collaboratively, instead of only standalone as “personal” assistants, and is relying on agents to do so. 

“Agents are where the value prop is,” said Directions on Microsoft analyst David Berry.  “They are purpose-built, narrow focus like mobile apps and it’s easier to build a business case around with higher visibility and quantifiable returns in many cases.  This makes them easier to point to and understand than an LLM (large language model).”   

“It’ll be interesting to see if Microsoft keeps including them in the M365 license or move some out to the PAYG (Pay as You Go) model,” Berry added. 

The company already rolled out its Researcher and Analyst agents and made them available for no additional charge to Microsoft 365 Copilot subscribers but limits a user’s combined queries to 25 per month. It also recently rebranded and repositioned its Finance, Sales and Service “role-based” Copilots and added them to Microsoft 365 Copilot subscriptions for no additional charge (rather than the $20 per user per month additional fee that some of them cost previously). 

You Get an Agent… and You… and You 

If you (like our analysts at Directions) thought keeping up with the Microsoft Copilot proliferation was challenging, buckle up for the agent onslaught. 

Microsoft is advocating that there should be an agent for every Teams meeting, specifically the Facilitator Agent. Facilitator will be able to do things like create an agenda, take real-time notes, capturing and assigning tasks, creating documents, capturing ad-hoc discussions. Facilitator is now available to those with a Microsoft 365 Copilot license; the task management and documentation creation components are still just in public preview. 

Microsoft also is pushing the idea that there should be an agent for every Teams channel. These Channel Agents can create status reports for projects; provide answers based on the channel’s previous conversations, meetings, plans; and stay up to date on the status of action items. Channel Agent is in public preview for those with a M365 Copilot license, with general availability scheduled to kick off in Jan. 2026. 

And there also should be an agent for every community, in Microsoft’s world view. Viva Engage community members can test the Community Agent, meant to assist with proactive support, creating responses, and controlling posting responses. The Community Agent in Viva Engage is in public preview for those with a M365 Copilot license. 

SharePoint is getting a Knowledge Agent, now in public preview, and due to start rolling out in Jan. 2026. Knowledge Agent can assist with content management, tagging and classifying files with metadata, improving sites and helping organize libraries. And a Project Manager Agent for Planner and Teams is coming, as well. This agent already is in public preview for those with a M365 Copilot license for Teams Meetings and Channels. As the name suggests, this agent helps create and keep track of plans and tasks. 

In late August, Broadcom announced (yet another) change to how they license VMware Cloud Foundation (VCF), the suite that includes VMware vSphere, vSAN, NSX, and cloud management software. That change goes into effect next month, on Oct. 16^th. 

VMware customers have traditionally had a license portability entitlement, allowing them to move their licenses between datacenters and even to other providers of VMware hosting. At the same, third-party VMware hosters like Microsoft’s Azure VMware Solution (AVS) and Amazon Elastic VMware Service (EVS) could also resell VCF licenses to their customers, allowing them to sell a comprehensive service. 

This latest change moves VCF to what Broadcom calls a “license portability-only operating model.” This means hyperscalers like Microsoft and Amazon can’t bundle VCF licenses in their offerings, and customers have to purchase licenses directly from Broadcom (or its reseller partners) to bring with them in a “bring your own license” (BYOL) scenario. 

Microsoft: A Less Safe Haven as of Oct. 16 

Since 2024, Microsoft has been promoting AVS as a safe haven from Broadcom’s licensing changes and price increases, even offering discounts and price guarantees for customers willing to make commitments of one, three or five years called reserved instances. Customers that have taken advantage of reserved instances won’t see any changes until their term expires, and of course Microsoft urges customers who want to extend their commitment to contact their sales team. But for new AVS customers or existing customers adding additional AVS nodes, making AVS purchases after Oct. 16^th means buying VCF licenses from Broadcom instead of getting everything from Microsoft. 

This is also a play by Broadcom to have a more direct role in the customer relationship, ensuring their sales team is involved with the end-customer even in hosted managed services like AVS. It also makes it easier for Broadcom to change (increase) pricing in the future, bypassing hosters like Microsoft and Amazon who might be able to exert more pressure due to their scale. 

Microsoft’s own announcement assures customers “These updates are about licensing only—there are no product changes to how Azure VMware Solution works. Microsoft will continue to deliver Azure VMware Solution as a fully managed VCF private cloud service on Azure.” The five-year AVS reserved instance Microsoft’s been offering means AVS will continue at least that long. And the BYOL version of AVS, already available worldwide, “is priced lower than AVS with bundled VCF subscription” although, at this writing, pricing is not yet available on the public AVS pricing page. 

But this may cool what has been a cordial relationship between Microsoft and VMware in recent years, even after the Broadcom purchase. Microsoft’s already more aggressively positioning Azure Local as a competitor to VMware on-premises (with recent updates to Azure Migrate easing migration from VMware to Azure Local), and AVS has always had the underlying motive of getting customers into Azure by any means possible.  

With Broadcom’s latest move to insert themselves into the AVS customer relationship, and Microsoft’s increased willingness to set aside anything that doesn’t immediately benefit the company’s focus on AI, AVS could have a more limited future once the current batch of reserved instance customers nears the end of their terms. 

This week, Microsoft announced it is rolling out Copilot Chat and web-grounded agents to the major Office applications.  Given the naming confusion around the Copilot brand, the significance of this may not be entirely clear to everyone. In short, what’s happening is Microsoft 365 subscribers will get Copilot Chat in Word, Excel, PowerPoint, Outlook and OneNote —even if they aren’t among those covered by M365 Copilot licenses costing $30 per user per month. 

Although M365 subscriptions jumped this year, which Microsoft attributed in part to M365 Copilot, Microsoft is still looking for ways to drive even more of the add-on M365 Copilot licenses. With its self-professed goal of a Copilot on every device, Microsoft arguably sees anything less than 100% adoption by its clients as unacceptable.  

What’s In the Box? 

What do you really get though, with chat in the apps?  Users not licensed for M365 Copilot will be able to open Copilot Chat in a task pane on the right side of the app. This gives you access to the chat interface you know from the M365 Copilot app and website.  Copilot Chat is tuned to respond based on that app you’re in. So, Copilot for Word will be focused on helping with documents, Copilot for Excel on data analysis, etc.   

What you don’t get are the in-app Copilot companions and builders, indicated by an AI icon, that come with the M365 Copilot license.  If you don’t have them, you may not miss them and when you do have them, you may find them annoyingly underfoot.  However, they can be useful once you get used to them by providing AI-assisted actions, like drafting and revising text using natural language prompts, using auto-rewrite, or creating a summary of the content. Each application has a slightly different implementation of the companions and have actions related to the specific app. 

What Can It Do? 

The unlicensed Copilot Chat in Office apps is still constrained to web-grounding, so there’s no access to work data via the Microsoft Graph.  Users can upload work files to the chat though to provide business context or use the built-in Context IQ which enables you to use a “/” in the prompt to open a file picker dialog with files relevant to you.  Adding business context is practically the great equalizer with the licensed version though, since this is one of the better ways to draft new Office files.  (And, in fact, Copilot for Word does not draw on business data in the Graph to create new documents anyway.)  

With Copilot Chat, you also have access to web-grounded agents but accessing business-grounded agents still require the pay-as-you-go model (PAYG).  And some agents, like Microsoft Researcher and Analyst, are not available as PAYG. They are available for no additional cost only to M365 Copilot licensees. 

What Does It Mean? 

Ultimately, the unlicensed Copilot Chat brings a boost of AI-powered productivity to Office apps, providing app-aware assistance that can streamline common tasks for all Microsoft 365 users. While its capabilities are limited compared to the full M365 Copilot, it still delivers practical value, especially for those not ready or able to invest in the full version. For many, this may be enough—at least for now. Yet, it’s clear that Microsoft is steering users toward the full benefits of the licensed tier, making this feature a likely gateway for broader adoption later. 

Microsoft is rebranding and repricing some of the products it had been selling as individual Copilots beginning in Oct. 2025. 

Microsoft is renaming its Sales, Service and Finance Copilots, as it announced in a blog post on September 10. It is now calling these three copilots “Microsoft 365 Copilot for Sales,” “Microsoft 365 Copilot for Service,” and “Microsoft 365 Copilot for Finance.” 

Microsoft had launched its separate Sales, Service and Finance Copilots over the past couple of years. It had priced these as US$20 per user per month add-ons to Microsoft 365 Copilot, itself priced at US$30 per user per month.

The Verge reported last week that Microsoft was planning to make these three Copilots available as part of Microsoft 365 Copilot for no additional charge. Currently, Microsoft’s pricing pages do not reflect pricing changes. But that’s going to happen next month, a company spokesperson confirmed. 

“The role-based solutions previously known as Copilot for Sales, Service, and Finance will be available for all Microsoft 365 Copilot customers, at no additional cost, starting mid-October. There will be no changes to functionality for existing role-based Copilot customers,” the spokesperson said. 

Don’t Blink or You’ll Miss the New Branding 

In recent weeks, Microsoft has been modifying the branding of some of its Copilots in its documentation, changing the names of several former Copilots to “agents.”  For example, Copilot for Finance, which never exited preview, was rebranded as “Finance agents” in March 2025. Now, Microsoft seems to be rebranding it again as “Microsoft 365 Copilot for Finance.” 

Until this week, Microsoft also had begun rebranding its Copilot for Service as “Microsoft 365 Copilot Service Agent.” But now that branding seemingly has changed to “Microsoft 365 Copilot for Service.” 

In case you’re wondering if this new wave of Copilot-focused naming means Microsoft is slowing down the agent-buzzword train, the answer is no. The Microsoft 365 Copilot Sales, Service and Finance offerings are going to be available in Microsoft’s Agent Store – which Microsoft also is now calling the “Microsoft 365 Copilot Agent Store.” 

Microsoft also seems to be doing some similar rebranding around Viva, based on a Microsoft Support page listing. The Microsoft Viva Engage Copilot is no longer in active development. Microsoft is now working on adding features to “Microsoft 365 Copilot in Viva Engage,” and dropping all work on Engage Copilot v.1. There’s been no word so far if Microsoft also is planning to add its various Viva Copilots to Microsoft 365 Copilot at some point.

Last week, I asked Microsoft if the company was trying to clean up its Copilot branding by focusing on its three main Copilots: Copilot (the consumer-focused Bing Chat-based Copilot), Microsoft 365 Copilot, and GitHub Copilot. A company spokesperson told me Microsoft “had nothing to share at this time.” 

As Directions on Microsoft has pointed out over the previous year, Microsoft’s many Copilot-branded offerings are not based on the same core technologies, nor do they pull from the same datasets.  Microsoft still has not shared Microsoft 365 Copilot sales numbers, leading some to wonder whether its relatively high price has hindered corporate adoption.

“If nothing else, it’s good to see Microsoft realizing the chaos the ‘single brand’ of Copilot has created for this fleet of disparate and disconnected services. Hopefully this is the beginning of a trend to harmonize the branding—and ideally the licensing—of Microsoft’s broad range of Copilots,” said Directions analyst Wes Miller.

By Barry Briggs

Recently OpenAI open-sourced two Large Language Models (LLMs); seemingly moments later, those models were made available on Azure AI Foundry and on the AI community website Hugging Face.

What are open-source language models and what do they mean for enterprise usage and applications? And what is driving the urgency some feel in releasing open-source models?

Weight a Minute!

LLMs, at their core, are based on multi-layer neural networks. Each little circle in the (highly simplified) diagram below represents a “neuron,” a small amount of code; typical LLMs have hundreds of billions or even trillions of these – which is why highly parallelized GPUs are well-suited for LLM workloads. The key element of neurons is the “weight:” how much importance the model places on the result of each calculation in each neuron.

Think of the inferencing process as solving a large mathematical equation with billions of variables each “weighted” by a number, where A, B, and C and so on are the weights:

Here’s the point: the weights are the secret sauce, somewhat analogous to the source code, of LLMs, the core IP – and thus the models behind OpenAI’s ChatGPT were closed-source, specifically meaning the weights used in the 175-billion parameter model were not disclosed.

(There’s much, much more to it of course: current LLMs owe much of their efficacy to an epochal paper published in 2017 called “Attention is All You Need,” which described how data is consumed and transformed in models – hence “Generative Pretrained Transformer” or GPT. But I digress…)

Customizing Models

Enterprises quickly realized that LLMs could be incredibly powerful and useful within their applications and environments. But how to safely connect or integrate an LLM with potentially sensitive enterprise data?

RAGs

One of the first, and still perhaps the most prevalent approaches is called RAG, or Retrieval Augmented Generation. RAG apps, as they’re termed, connect models to data via search techniques.

Here a model (hosted within the organization’s security boundary, such as an Azure tenant) parses the user’s question, extracts relevant terms, queries the connected HR application, and then formulates a natural-language response.

RAG apps are conceptually straightforward, relatively easy to create, and, if properly developed, secure. They are particularly useful for rapidly changing data such as news. However, the search process (which can involve multiple steps) can introduce latency and if the retrieved information is long, such as a multi-megabyte PDF, it can add expense (since costs are based on token count) and even overflow the LLM’s context window.

Fine Tuning a Closed Model

Fine tuning allows organizations to adjust the internal operation of the LLM itself. In fine tuning, developers “feed” enterprise data to an LLM (against, if using Azure, hosted inside the organizational tenant) having the effect of changing the LLM’s weights. That in turn (typically) creates a new instance of the model.

Fine tuning a closed-source model involves uploaded text or images in specific formats to the model via an API. It’s useful for scenarios in which specialized terminology such as medical or legal terms are needed, or enterprise-specific jargon such as product names and descriptions, equipment manuals, scientific writing, proprietary terminology, and in general in scenarios where the data does not change frequently. Morgan Stanley uses a fine-tuned LLM based on GPT-4 to assist its financial advisors with complex questions.

However, fine tuning closed models can incur costs – invoking the fine-tuning APIs can be expensive, and inferencing fine-tuned models generally comes with additional cost.

Finally, and perhaps most importantly, closed models such as GPT-4o cannot be hosted locally, that is, on-premises, precluding a number of important use cases: meaning that their use incurs cloud costs.

The Emergence – and Importance — of Open-Source Models

Open-source models are freely available, downloadable, and, most importantly, expose their weights directly for fine-tuning. In the last few years we’ve seen, perhaps somewhat surprisingly given the investment required to create them, numerous “open-source” models appear.

Open-source models can be fine-tuned (a process not dissimilar to training) locally on customer hardware – even on your PC – so that experimentation, testing, and deployment can happen inexpensively. Software encodes the input data (say, reference material of some sort) and then adjusts the weights of the model.

Thus, for example, dozens of fine-tuned LLMs have cropped up in nearly every aspect of healthcare and medicine (see a survey here); one, called Clinical Camel, uses a fine-tuned version of Meta’s LLaMa-2 model as a diagnostician, with impressive results. Other use cases include LLMs in edge computing, financial fraud detection, smart cities, eLearning, and manufacturing, among others.

While OpenAI published arguably the first open-source model, a reduced version of GPT-2 in 2019, it has since fallen behind: a tsunami of open-source models, many from China, have emerged. Indeed, when open-source DeepSeek (from Chinese firm High-Flyer) launched earlier this year with performance comparable to OpenAI’s latest models, markets briefly tumbled.

Which is why, finally, OpenAI’s release of open-source GPT models is significant, and why Microsoft was quick to incorporate them into Azure AI Foundry. Note, in the chart above from the (extraordinary) benchmarking site Artificial Analytics, that prior to gpt-oss’s launch, the top right quadrant would have been primarily populated with Chinese models: Alibaba’s Qwen, Minimax, DeepSeek, and ZAI (GLM).

The Geopolitics of It All

As early as 2017, China’s State Council declared AI a national priority. Many believe its development and promulgation of open-source models enables “soft-power” politics in developing countries that would otherwise be unwilling to pay for higher-priced models.

And the efficacy and popularity of these models has (some would say finally) gotten the attention of American government and industry. In rapid order came the White House’s “AI Action Plan” on July 23^rd, the release of OpenAI’s gpt-oss models to great hoopla just days later, Microsoft’s announcement of their availability on AI Foundry, and the creation of the ATOM Project (American Truly Open Models) shortly thereafter (full disclosure: I am a signatory to ATOM).

Open, But Not Necessarily Free

(Microsoft sources: gpt-oss; gpt-5.
*Input charges may be less if prompts reused or cached.)

Now, squirreled away at the bottom of Microsoft’s announcement is the pricing of the gpt-oss models: that’s right, on Azure, the open-source models are not free! The larger of the two models costs $0.15/1 million input tokens and 60 cents per million output tokens; the smaller model, which can run on a single VM, is charged based on the VM cost. But – and here’s the deal – as we’ve said, these models can be downloaded and run for free on customer hardware.

Of course, OpenAI and Microsoft want to maintain some differentiation between open-source and premium, paid-for models; GPT-5 (on Azure, nearly 10x as expensive as gpt-oss, and not downloadable) supports many more features, has been extensively red-teamed, and is integrated with AI Foundry’s model router, a cost-saving feature.

What to make of all this?

A while back, I made the claim that models are increasingly commoditized. And with the emergence of open-source models customers can fine-tune them to meet the specific needs of their scenarios.

All this means there’s lots of choice, and customers will have decide which upon which criteria they choose their models. Will it be on the basis of cost? The ability to download them? The use of Responsible AI guardrails? National origin?

As is the case so often the answer is: it depends.

Think I need some fine-tuning? Drop me a line at bbriggs@directionsonmicrosoft.com.

No models were harmed in the creation of this blog post.

Microsoft has been working to move more customers off Enterprise Agreement (EA) volume licenses for the past few years. But it is really stepping up the war on the volume discounts on which these customers counted starting in Nov. 2025. Starting then, Microsoft will be discontinuing existing discounts for online services provided across all of its EA tiers, as well as for products covered by the Microsoft Products and Services Agreement (MPSA). 

Volume-licensing customers of all sizes, especially larger customers earning larger discounts, could be hit with substantial price increases as a result. In practice, the biggest and most strategic Microsoft customers already are likely getting better deals than the built-in volume discounts by negotiating with Microsoft. But the change will likely mean that fewer customers will be able to count on guaranteed discounts, leading to more budget pressure. 

Microsoft announced on August 12 plans to flatten these programmatic volume discounts across its online services — including Microsoft 365, Dynamics 365, Windows 365, and all security, compliance and identity management products. On-premises software products are not affected, and education and federal/state/local government customers are exempt. 

Before this change, volume customers typically received discounts of anywhere from 6% to 12%  on their online services purchased via EAs, with those purchasing more getting the biggest discounts. “Level D” customers which have 15,000+ users have received the automatic 12% discount on their EA services purchases. The coming Microsoft discount change – which we at Directions on Microsoft already are calling “Microsoft’s One Big Beautiful Licensing Announcement” — will affect Level B through D SKUs. Listed prices will be adjusted to match the Level A list price, which is the same as the Microsoft website price, according to Microsoft.  

Microsoft’s official reason for the coming change in services discounts: “This update is part of Microsoft’s ongoing efforts to simplify licensing and improve pricing clarity for our customers.” 

“With a standardized Online Services starting price for all customers across both Enterprise Agreement and Cloud Solution Provider, partners can invest time in pitching their value-added services instead of explaining the complexities of pricing structures,” explained Microsoft in an update to its volume-licensing hub on Microsoft Learn. 

It’s All About the Money 

But neither customers nor licensing experts buy that explanation. Instead, they say the move is part of Microsoft’s ongoing push to move all but the largest commercial customers from EAs to other types of channels like Microsoft Customer Agreements (MCAs) and MCAs for Enterprise (MCA-E).  

Microsoft has been actively working to require smaller and mid-size customers to work with Cloud Solution Providers (CSPs), while grabbing the largest volume customers for itself. The thinking: By eliminating volume discounts and taking the biggest customers direct, Microsoft will be able to grow its services revenues faster. 

For EA contracts which started on or before Oct. 1 2025, the no-programmatic discount rule kicks in only when, subsequent to Oct. 31 2025, customers add an online service that they had not previously purchased. For all EA contracts started on or after Nov. 1, the no-programmatic discount applies for all purchases. 

In 2017, Microsoft dropped Azure price level discounts, and in 2018, it eliminated the volume discounts for its smaller volume-discount customers (known as Level A).  

Microsoft’s rounds of layoffs during the first part of this year already meant that some customers lost their dedicated Microsoft account managers, noted Directions on Microsoft‘s Rob Horwitz. And the coming volume-licensing changes are going to make the situation even more confusing.

“Some programmatic discounts are being eliminated. The names of some things are changing. But the importance of and procedures for preparing for an upcoming contract negotiation remain largely unchanged ,” Horwitz said. “For example, it is still important to assemble and manage your internal contract negotiation team, clarify your current use and future needs for Microsoft products, understand the special concessions that you got last time that must be maintained, and develop financial models for various scenarios.”

And that’s where an advisory service like Directions on Microsoft can be invaluable to companies who are going to have to learn to navigate this new Microsoft licensing landscape, he added.

Microsoft beat its own and Wall Street’s expectations with its Q4 FY25 earnings, and the company’s stock price and market cap hit all-time highs this past week. Nice if you’re a shareholder or employee who survived the latest rounds of layoffs. But if you’re a Microsoft enterprise customer, what’s your takeaway from all this? 

There are some clues in the numbers that Microsoft did — and didn’t — disclose in its Q4 report. 

Azure, By the Numbers 

For the first time ever, Microsoft provided an actual Azure revenue number as part of its Q4 report. For years, the company has declined to do so, speaking only in growth percentages without disclosing revenues, seemingly to prevent itself from being compared to AWS and Google Cloud. 

Microsoft’s overall revenues for FY25 were $282 billion. Azure revenues were slightly less than half of the $168 billion in “Microsoft Cloud” revenue the company measured for fiscal 2025. (“Microsoft Cloud” is Microsoft’s own metric consisting of cloud revenues from Azure, Microsoft 365, Dynamics 365, some components of GitHub and LinkedIn and more.) 

For its fiscal 2025, ending June 30, 2025, Azure and “other cloud services” surpassed $75 billion in revenue, up 34 percent year-over-year, Microsoft said, and it’s estimated to grow another 37% in Q1 FY26. Comparatively, AWS is at an annual revenue run rate of more than $123 billion, and Google Cloud at more than $50 billion. 

Oddly, Microsoft’s latest earnings report didn’t disclose how much of its Azure revenue was attributable to AI. During the past several quarters, Microsoft provided either a number of points or a percentage of Azure revenues that came from AI workloads. Officials didn’t say why they decided against providing that figure this time around. 

Microsoft execs instead emphasized that Azure revenues in FY25 were attributable across “all workloads.” They said Microsoft now has more than 400 datacenters across 70 regions, and that cloud demand remains higher than supply. In Q4 alone, Microsoft’s capital expenditures were $24.2 billion, with most of that going toward datacenter leases, servers, GPUs and other infrastructural components. Microsoft plans to continue to grow capex, predicting Q1 FY26 capex spending would hit $30 billion.  

Microsoft 365: Numbers Shared and Not 

Meanwhile, Microsoft 365 commercial had a strong Q4 as well, with revenues up 16%. Microsoft officials attributed seat growth primarily to new small-and-midsize businesses (SMBs) and frontline workers. They attributed growth in revenue per user to high-end E5 subscriptions, as well as Microsoft 365 Copilot. 

Microsoft has not released an updated number as to what percentage of the Microsoft 365 commercial base is on E5 in years. In 2022, that number was just 12%. Each year since then, Microsoft sales has made migrating and upselling customers to E5 a priority.

Microsoft also has yet to release a number as to how many Microsoft 365 Copilot subscriptions it has sold. Instead, like it used to do with Azure before this quarter, Microsoft execs continue to provide difficult to quantify numbers about Copilot sales and adoption. 

For example, officials said the company now has 100 million monthly active users “across our family of Copilots” in both the commercial and consumer space, but it’s not clear if this figure includes M365 Copilot Chat, which is free for business customers with M365 subscriptions. Microsoft also said GitHub Copilot now has 20 million users and the Dragon Copilot had 13 million “physician-patient encounters.” They added that Microsoft 365 Copilot saw “the largest quarter of seat adds since launch,” in Q4 FY25, and a record number of customers returning to buy more seats. But there was no actual Microsoft 365 Copilot number or any talk of revenues coming from M365 Copilot during the earnings call. 

Reading the Earnings Tea Leaves

There are some takeaways from these numbers (and non-numbers) that are likely to affect enterprise customers in the coming year.  

First, Microsoft’s pressure to adopt its various Copilots won’t be letting up. It’s a safe bet to assume adding more commercial Copilot seats and licenses will be a key part of any Microsoft licensing negotiation.  

Ditto with the priciest Microsoft 365 tier. Microsoft and partners will continue their mission of getting even more customers to go with E5. Those who already are there should expect Microsoft to be pushing paid add-ons to their E5 subscriptions for things like Teams Premium, SharePoint Advanced Management (for the entire company, not just M365 Copilot users), etc.  

Company officials noted on the earnings call that Microsoft Purview data governance and security tools are already used by three quarters of Microsoft 365 Copilot customers, but there’s still room for more add-ons there, too. 

“Customer adoption of Copilot and AI is Microsoft’s biggest priority,” said Directions analyst Jim Gaynor. “And there will be the need for even more add-ons, as organizations realize they need things like Purview, SAM, and Azure AI Content Safety to complete the picture. Microsoft’s relying on that to keep increasing their revenue.” 

Remember the introduction of ChatGPT on that momentous day in November 2022? Since then Copilots and chatbots have sprung up everywhere, social media is flooded with AI-generated posts, and enterprise-class agent-based applications are all the rage among techies and analysts. What an extraordinary time! 

I was asked recently where AI, and for that matter, enterprise computing will be in five years. With the rapid pace of change it’s obviously more challenging than ever to say – but after some thought here’s where I came out. 

No warranties or guarantees given or implied!  

The Conservative View

I provided two possible views: a “conservative” one, in which AI is gradually and incrementally absorbed into the enterprise ecosystem, and a “radical” one in which AI changes everything.  

The conservative view holds that, over time, applications absorb AI functionality, and we come to expect AI in all of them; in short, AI capabilities become an application convention just like, say, the Office ribbon or Copy/Paste. Copilots gradually improve. They connect to every application and data source in the enterprise in a safe, secure fashion (and as I describe in my discussion on data governance, protecting your data could in fact, occupy all five years!). Copilots live not just in Office but everywhere; perhaps Microsoft will even offer a “Copilot certification” for third-party apps to use its libraries and models and the Copilot trademark. 

Less rote, pedestrian code is written by humans but, conversely, a higher premium is placed on developers and architects who can piece together the Copilot or Claude Code or Gemini code fragments into ever higher- and higher-value applications. Profound coding knowledge is still essential, but more and more developers check AI-generated code – rather than write it. 

In this view, autonomous LLM-powered agents achieve a certain level of acceptance but (like Lotus Notes Agents, introduced way back in 1993) compete with existing and entrenched RPA apps. Moreover, users and enterprises, perhaps rightly, have limited trust in agents to perform mission-critical functions given their nondeterministic, statistical, probability-based nature and their tendency to hallucinate. 

The Radical View 

AI in this view changes everything in the next five years: we won’t recognize the new world of computing. 

Imagine, for a moment, a contact center application completely comprised of agents.  

In this completely-made-up example, a Channel Agent manages the various customer contact channels (phone, email, etc.), perhaps assigning different priorities to regions at different times of the day. A Customer Analysis Agent identifies the customer and perhaps determines the likely cause of the call; a Routing Agent sends the call to either the agent or human most likely to successfully resolve it; and so on.  

These agents are continuously trained on the vast amounts of (now well-governed) data in the enterprise, allowing them to quickly build and adapt data models, identify emerging trends and adjust business processes accordingly – and in general relieve IT of much of its traditional tedium. 

IT Development in Flux

In theory, “developers” of the future create no procedural code; coding becomes a lost, or rare, art. Because all the agents are programmed, in effect, in English, the application can be completely customized to the very specific needs – indeed, nuances – of the organization, and (again in theory) could be easily changed as business needs mandate. There are no packaged ERP or CRM applications – rather, vendors provide infinitely customizable agents which IT teams with “SDKs” describing how to create the appropriate prompts. 

Nevertheless, IT technologists remain a valuable and scarce resource. Profound knowledge of the inner workings of LLMs, the mechanics of exposing data safely and connecting LLMs to them, keeping these new ecosystems of agents and data secure, and creating the necessary goals and reward systems to achieve business goals: all these valuable new skills power the new IT. 

What About Office? 

Remember: the only reason Excel, PowerPoint, and Word are separate applications is because they were created in a time when PC memory maxed out at 640 kilobytes. Now most enterprise laptops sport (minimally) 16 gigabytes, or over 26,000 times that original, oh-so-primitive capacity.  

Today, how many of you have all three open right now? (I do.) And we still need to know the silly, arcane, and dated methods of cutting-and-pasting between them. (Edit / Paste-Special? Really?) 

So perhaps all the metaphors upon which we depend today undergo a growth spurt: with no need to respect hardware limitations, and LLMs’ extraordinary ability to quickly ascertain user intent, we no longer need separate applications for word processing, presentation graphics, and numerical analysis all possessing unintuitive menu hierarchies, quirky ribbons and distinct UIs. 

In other words, could the Office apps dissolve and be replaced by…something else? A unified, all-in-one application for content creation, presentation and dissemination? (Maybe we’re already seeing the beginnings as OpenAI and Perplexity (and even Microsoft, sort of) introduce their own AI-native browsers).  

Which Will It Be? 

It’s my assertion that the mission of IT, ultimately, is to give control of the business to the business; that is, to enable, in real time, digital systems to respond to business leaders’ direction. Why couldn’t a CEO simply speak into some UI, “Decrease the price of our Widgets in Japan by 5% tomorrow – and provide an analysis the next day showing which market demographics and segments responded?”  

So, which will get us there faster: incremental evolution or a revolution in IT? Of course, I don’t know – but I suspect we’ll see a continuing tension between a natural IT conservatism (if it works, don’t break it) set against the rapid, nonstop acceleration of AI – and the overarching necessity of driving competitive advantage.  

Scary, fun, or just plain crazy? Drop me a line at bbriggs@directionsonmicrosoft.com and let me know what you think.  

Microsoft recently announced SharePoint Advanced Management (SAM) was rolled into M365 Copilot licenses for no additional cost.  Originally, Microsoft licensed SAM separately, so this is a big advantage for organizations wanting to roll out M365 Copilot.   

Or is it?  

Although Microsoft touts SAM as helping prepare organizations to deploy M365 Copilot, its tools are intended to identify and potentially resolve problem areas in SharePoint and OneDrive for Business and, consequently, is supposed to be licensed for everyone.  As a result, this “no cost” pitch is creating confusion among customers.  If organizations want to use SAM for rolling out M365 Copilot, does everyone in your organization need a M365 Copilot or a SAM license? 

Microsoft is working to drive adoption of M365 Copilot, its flagship AI service, and knock down barriers to deploying it for its customers.  Part of that strategy appears to be coupling SAM with M365 Copilot so that when an M365 Copilot license is purchased, SAM is enabled for the customer’s tenancy by default. This may not be a bad thing technically since SAM provides several tools for identifying and remediating high risk areas including oversharing, permission sprawl, site sprawl and outdated content. However, these risks also impact enterprise search as well as SharePoint in general.  M365 Copilot just happens to amplify their exposure. 

Since SAM’s capabilities apply to enterprise search and SharePoint (after all, it’s SharePoint Advanced Management, not Copilot Advanced Management), it would seem to support the idea that SAM licenses should be required for all users of SharePoint and OneDrive for Business.   

Ultimately, all customers are governed by Microsoft’s Product Terms. When you read the Product Terms and SAM product documentation, there is no indication that the standalone SAM User SL (at US$3 per user per month) is waived for anyone other than people with M365 Copilot licenses.  Consider an organization that buys M365 Copilot licenses (at US$30 per user per month) for 10% of its employees. Having to license SAM for the remaining 90% of its employees could double its expenditure.    

Where the confusion comes in is with Microsoft’s public blog posts and account team conversations with company representatives using language like “no cost” and “tenant-level service” for SAM. This language can be misleading (and has no legal standing anyway).  

Based on current information, we believe this “no cost” language appears to refer only to licensed M365 Copilot users, and “tenant-level service” does not equate to a tenant-level license (as they don’t even exist).  The result: When you start using SAM to prepare for M365 Copilot, you may be exposing your organization to a licensing violation since — in the future– Microsoft could argue your non-licensed M365 Copilot users benefited from the SAM tools and must be licensed. 

What Are Your Options?

If you decide to deploy M365 Copilot, what are your options related to SAM?  

You may want to get written confirmation from your Microsoft account manager that, given your use of SAM is for the deployment of M365 Copilot, that SAM User SLs are not required for unlicensed M365 Copilot users.  However, while this may offer you some protection, you still run the risk that Microsoft may decide to enforce the licensing at your next renewal.  Alternately, you can evaluate a third-party product (also a cost) or inform your administrators not to use SAM, being aware that SAM capabilities can be auditable.  

Often Microsoft licensing is convoluted. I don’t think anyone at Microsoft is intentionally trying to trap customers into buying SAM as much as they are driving M365 Copilot usage.  However, the fact remains, the current Microsoft Product Terms (and SAM documentation) say every user who accesses a licensed online service must have a User SL.  And it is not uncommon for Microsoft to take a more laid-back view of licensing rules while it tries to get a technology adopted by customers, only to embrace a stricter interpretation once the technology becomes mainstream.  

By Barry Briggs

Let me start with a perhaps-controversial statement: It’s my opinion that every Chief Technology Officer and, for that matter, every Enterprise Architect, every Solution Architect, and every Information Architect should code. Coding is the only way that you can stay grounded in the fundamental, ultimately rigidly binary, logic of how computers work; what their limitations are; and what the limitations of humans writing code are. Frankly it’s easy to draw a block diagram of boxes and arrows on the whiteboard; translating that into working, secure software is an entirely different matter altogether.

Most importantly: To effectively manage dev teams, we have to understand what they’re doing, how they’re doing it, and the challenges they face.

These days the buzz is about “vibe coding,” meaning using AI to write code, maybe even whole programs. And as the Wall Street Journal informs us, even professional developers use it now. 

Is vibe a revolution in coding? Can it help a longtime programmer? And how should executives think about it?

I’m a professional developer. And a six-time CTO. I decided to find out.

The Problem

Hopefully you’ve all noticed that we at Directions have upgraded our website . (And if you’re not a subscriber: you’re really missing a lot of great content.) I wanted to do some analytics on the data and, long story short, I needed to extract it – all of it – from our back-end WordPress site.

We’ve all known for a while you can use GitHub Copilot and many other coding assistants to help with coding. Copilots are now available in Visual Studio, Visual Studio Code; and if you’d rather, Gemini, Perplexity, Cursor, and Claude Code can write code for you too. More on that later.

Like all coders, after a time I found myself “in the zone” with headphones on, totally focused, immersed in the logic, adding features, refactoring, discovering and handling exceptions, the world completely shut out. (It’s really kind of wonderful.)

But things have changed – bigly –  and if you’re a coder, or, more importantly, if you manage developers, you need to know.

The New Dynamic

Not that long ago if your code (yours, not mine) had an error, you spent time Googling various sites – Microsoft’s documentation, the developer problem-solving site StackOverflow, Reddit, and blogs – trying to find how to fix it.

Then Copilots arrived and you could ask them – maybe tentatively, as in the early days the results might be questionable.

But now the dynamic has reversed. As a good friend of mine noted, no longer are Copilots assisting us; we humans assist the Copilots! As an experiment, within the Visual Studio IDE, I simply asked Copilot to “create a console app, load all the JSON files in d:\files and parse them into a C# class.” Done. Then I asked it to count the number of entries with specific metadata tags, to get an idea of how our coverage was distributed.

Done. Boom.

It was fun; it was fast; but I began to wonder: was I really coding?

Hold that thought.

It wasn’t perfect. Occasionally Copilot wrote code that either didn’t work or didn’t compile, usually because it had found sample code somewhere that used a deprecated API, or whatever. Usually these were simple fixes, but it meant I had to type more prompts like “please fix line 366” and the like. Would I have gotten it right the first time? Hmmm.

Interestingly, I fed many of the prompts to other LLMs – Gemini, Claude, Perplexity – and in some cases their code was improved, while in others the code was, well, even worse.

And – cautionary note. In a project I was doing with Azure AI Foundry, I got the following error:

There was an error generating a response. Chat history can't be saved at this time. Error code: 401 - {'error': {'code': 'PermissionDenied', 'message': 'The principal XXXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXX lacks the required data action Microsoft.CognitiveServices/accounts/OpenAI/deployments/chat/completions/action to perform POST /openai/deployments/{deployment-id}/chat/completions operation.'}}

I had no idea what this meant – seemingly something about Managed Identities. I asked every LLM: Claude, Gemini, Azure Copilot, ChatGPT, Perplexity, GitHub Copilot, and received different answers from each one. And all were wrong. Great.

The Meaning of Vibe Coding

The real question for technology executives is – do AI assistants accelerate development? And can you replace developers?

My answer on the first one is: the jury’s out. On the second: categorically no.

Why?

• Copilots can save an immense amount of time on the minutiae of looking up API calls, parameters, and advising coders on their correct usage. And they can write functions that actually work, as I mentioned above. And fix simple bugs. That’s kind of awesome.
  
  
• But when the routine doesn’t work, you need someone who understands all that minutiae to dissect the code and fix it. For example, I ran into a minor problem with a Unicode string in the data. Copilot’s answer was misleading and its solution wrong. But because I understand Unicode and its use, I could fix it quickly. Point: you need experienced coders.
  
  
• A totally empirical and anecdotal finding – but for me important – was that while I increasingly relied on Copilot to do rote coding, I spent nearly as much time typing prompts as I would have typing code.

Vibe Coding in the Enterprise: Lessons for the CxO

To be clear, these days my apps are what we might call “quick and dirty.” Enterprise-class robust? Hardly.

Having sat through countless architecture and code reviews at Microsoft and elsewhere, I have a pretty good idea of what it takes to pass muster. Copilot-written code needs every bit as much scrutiny and review as human-written code – more, I’d say. Repeating the point: You need experienced programmers!

That said, the ability of the Copilots to connect to enterprise code repos, to leverage organizational coding standards, even to generate and run unit tests, are all remarkable but in no way obviate the need for skilled technical talent. Instead, you should expect more. If these tools amplify productivity, then set more ambitious goals. For your developers – and for your organization.

Have you vibe-coded? What do you think? 

I’m at bbriggs@directionsonmicrosoft.com. 

Some customers will be able to buy access to critical security updates for obsolete versions of Exchange Server and Skype for Business Server starting August 1, 2025. The just-announced Extended Security Updates (ESUs) programs for Exchange and Skype for Business reduce risk for customers migrating to Exchange Server SE and Skype for Business Server SE, which will soon be the only supported versions of those on-premises servers.  

The following versions are eligible for ESUs: 

• Exchange Server 2016 CU23 

• Exchange Server 2019 CU14 or CU15 

• Skype for Business 2015 

• Skype for Business 2019. 

All of those versions leave product support on Oct. 15, 2025. Microsoft had said it would not offer any support for these Exchange and Skype for Business versions after that date, and was strongly urging customers to treat Oct. 15 as an absolute (while admittedly tight) deadline for migration. Microsoft is still urging but also hedging by offering access to critical security updates through Apr. 14, 2026, via ESUs. 

How the Exchange, Skype ESUs Are Different From Other ESUs 

Exchange and Skype ESU licenses and media will not ship the same way as Windows, Windows Server, and SQL Server ESUs. Instead, Exchange and Skype for Business customers must apply for the ESU programs through their Microsoft account team (who will provide pricing details) starting Aug. 1, 2025. Microsoft does not specify which customers are eligible, but they must not be able to finish migrating before Oct. 15. And they must be big enough or important enough to have a Microsoft account team assigned. 

Importantly, these are brief ESUs – only for six months – and there’s no guarantee there will be any Critical or Important security updates to release; the program is “just in case” insurance. Also, Microsoft claims the ESU period won’t be extended, saying “you don’t need to ask.” But they were adamant that support for those older versions would end in Oct. 2025, too. 

Related Resources

Announcing Exchange 2016 / 2019 Extended Security Update program. 

Announcing Skype for Business 2015 / 2019 Extended Security Update program. 

After much hype, Microsoft 365 (M365) Copilot Search is rolling out. Part of the set of M365 Copilot Wave 2 spring release announcements, M365 Copilot Search will be available to M365 Copilot licensees by August 2025.   

M365 Copilot Search is optimized to find specific content items and can do it without knowing precise terms or keywords.  It’s the next evolution of a universal search tool Microsoft has been trying to land since 2017 to span across SharePoint, Office apps, email and more, intending to keep users in their flow.   

M365 Copilot Search extends the M365 Copilot product to provide contextual results to people writing natural language queries while respecting the security boundaries of their organizations.  This is potentially an improvement over the current Microsoft Search feature found in the top bar of SharePoint and the Office apps, which is more reliant on traditional keyword searches and can produce unsatisfactory results. Unlike M365 Copilot though, it does not leverage Bing search. 

How M365 Copilot Search Works 

M365 Copilot connects to content in your organization through the Microsoft Graph and uses semantic indexing to map relationships and contextual information to locate relevant files for you.  As a result, you don’t necessarily have to recall the exact title of that document you worked on to search for it with M365 Copilot Search as long as you can recall something relevant about the document or even who worked on it with you.   

At least to start, M365 Copilot Search will be available through the M365 app and website only to M365 Copilot-licensed users, making it unlikely it will replace Microsoft Search for the time being.  Its home page can almost act as a personalized dashboard (distantly related to SharePoint’s MySite concept) which some people may find useful as a starting point for their days.   

Aside from the Search box, M365 Copilot Search has two core sections loaded with content relevant to you: Recommended and Quick access.  

The Recommended section reflects recent documents with which you engaged as determined by M365 Copilot Search using a combination of Microsoft Graph, M365 Copilot and user context to determine what shows.  Each recommendation includes a couple of pre-defined prompts intended to help you work with the content.  There seems to be a fixed number of these prompts, and you might find their value, like Summarize recent changes and List key points, subjective based on the file. Clicking on one of the prompts will take you to M365 Copilot Chat.  

Quick Access categorizes your content into three self-described tabs: Recent, Shared and Favorites. M365 Copilot Search allows you to filter on not only the typical Office and PDF file types, but also Copilot chats, notebooks, pages and creator, for a current total of 20 file types.  All three tabs indicate when you last interacted with the file and can indicate your action.  

I’m not convinced how helpful these actions are yet. Some actions like “You frequently edit this” and “You recently opened this” aren’t particularly useful to me since I know what I’m working on, but actions indicating who shared or sent a file might be.  Inspired by the “Who can see this?” feature discussed below, I’d like to see this space also used to indicate a file’s sharing risks, including Anyone links, Everyone Except External Users and external (guest) users. Oversharing is a critical problem in many organizations and specifically when using M365 Copilot so this enhancement could help  fix it.  

Another AI enhanced feature in Quick Access is file summaries.  Hovering over a file list in Quick Access will pop up a dialog that generates a synopsis of its content, so you don’t have to open the file to check what it contains.  It may take a few seconds to generate it based on the file, and right now it seems to support mostly the Office file types.  Hopefully, these summaries will include Teams meetings and invites soon.  It’s interesting that this feature isn’t available in the Recommended section or in search results, or at least not yet. 

A Potentially Powerful Tool… But Is It Powerful Enough?        

Speaking of search results, the search results can be filtered and sorted like regular search engines and have nine default enterprise search verticals (filters), like Files, Sites and People.  M365 Copilot Search can use OCR to find text in images as well as look into transcripts of videos and meetings, and inside Power BI reports.  An intriguing feature on the context menu is “Who can see this?”.  “Who can see this?” opens a convenient Manage Access dialog box which displays the People, Groups and Links associated with that particular file. Given the risks of oversharing in the age of Copilot, this can be a really useful feature and I’m a little surprised it’s not available on the Recommended or Quick access context menus also.  

M365 Copilot Search blends traditional search with AI chat, appearing to offer a potentially powerful productivity tool for users managing an ever-increasing volume of content.  Like M365 Copilot, users will likely have to spend some time learning how to write their search prompts to get the most out of their queries.  Since Copilot Search relies on organizational data, users are required to have a M365 Copilot license. While this capability may be valuable, it alone may not be compelling enough for most organizations to bite.  

“Bet-the-business” strategies aren’t forever. A prime example: Microsoft’s years-long efforts to grow its consumer and enterprise mixed-reality hardware, software and services. 

Just a few years ago, in 2022, Microsoft CEO Satya Nadella was touting the idea that Microsoft could be a leader in multiple metaverses, from gaming to commercial. On the enterprise side, the company was positioning its HoloLens, Mesh, IoT services and digital twins as key components of its evolving enterprise metaverse stack.  

After a string of setbacks for the HoloLens (ranging from leadership issues to problems with a $22 billion deal for ruggedized HoloLens hardware for the U.S. Army), Microsoft announced it would completely discontinue support for its HoloLens 2 headsets of Dec. 31, 2027.  

Now it’s also dismantling Mesh, its set of technologies that enable shared 3D virtual environments. In 2024, Microsoft made Mesh generally available in Teams meetings so that users who weren’t physically co-located could share an “immersive” virtual space. The Mesh platform, which some likened to the virtual 3D world known as Second Life, never really caught on. Then “return to office” initiatives started undoing remote and hybrid work arrangements, and AI overtook the metaverse/Web 3.0 buzzwords on Wall Street and beyond.

Mesh’s ‘Evolution’ = Discontinuation 

In a message on the Microsoft 365 Admin Center in May, Microsoft officials told customers to expect Mesh to “evolve.” At that time, in MC1074973, Microsoft said it would be retiring the Mesh Toolkit effective June 24, 2025. At the same time, officials said: 

“Microsoft is excited to continue investment in immersive experiences, and we will continue to share more information about how Mesh is evolving,” 

A month later, customers learned how Microsoft defined “evolution.” Microsoft issued a new message (MC1111181) that noted Mesh was being retired and replaced with Teams immersive as of Dec. 1, 2025. Microsoft is pulling the plug on the Mesh PC and Quest apps, the mesh.cloud.microsoft website and the Immersive Space (3D) view in Teams as of that date. 

Confusingly, immersive events in Teams will remain, which are different from the Immersive Space view in Teams. Immersive 3D meetings in Teams are for small (up-to 16 people) casual meetings, scrums, standups, brainstorming sessions, etc. Mesh events were for up to 330 people for things like all-hands, town halls, onboardings, etc. Mesh events also let companies create interactive objects and highly customized 3D spaces like a virtual version of a company lobby, while Teams immersive relies on pre-created meeting “spaces”. 

“Mixed reality just didn’t become the big opportunity that Microsoft hoped at the time,” said Directions on Microsoft analyst Jim Gaynor. “Retiring Mesh looks like the last stages of winding down something that’s not a priority for Microsoft any longer as they keep pruning away things that don’t contribute to AI and Copilot.”  

Teams immersive is accessible via the Teams calendar. “To schedule an immersive event, select the drop-down next to New Event, then choose Immersive event,” Microsoft explains.

The progress of artificial intelligence in the last few years, we’d all agree, has been phenomenal. Five years ago, trapped in the misery of Covid, none of us foresaw the emergence of ChatGPT and its follow-on effects – that Microsoft would invest $13 billion in OpenAI, that Nvidia would be approaching a $4 trillion market cap, that tech vendors are scrambling to build data centers. If everything’s about AI these days, then capex truly has become strategy. 

But can machines ever approach the human capacity for intelligence?  

It’s not – at all – an abstract question. Because if OpenAI ever achieves Artificial General Intelligence (AGI), as it’s called, then, contractually, they are no longer bound by the exclusivity clause in their contract with Microsoft.  

If you’re Satya Nadella, that’s a big deal. Microsoft bases much of its market leadership in AI on its exclusive access to OpenAI models, which underly all those Copilots and Copilot Studio, and are available for use in Azure. 

What is AGI?  

A very good question, and one which has many possible answers. Some argue that it’s passing the famous Turing Test: which is, briefly, that conversation with a computer is indistinguishable from chatting with another person. By that measure, we may have already achieved AGI.  

Another test (called generality) involves how broad a range of topics an AI can be used for. An LLM trained on the entire internet can chat with you about much more than one trained just on your local enterprise data. Here again, we seem to be doing pretty well: a recent Harvard Business Review study showed that the top use case for AI in 2025 is…wait for it…therapy and companionship. Numbers two and three: “organizing my life” and “finding purpose.”  

Check it out: 

Top 10 GenAI Use Cases  
Harvard Business Review 

In 2023, Mustafa Suleyman (now CEO of AI at Microsoft) and Michael Bhaskar (AI strategy and communications at Microsoft) suggested a more quantitative test: give an AI $100,000, let it invest, and turn that sum into $1,000,000. Such a task involves complex reasoning, analytics, and forecasting.  

We’re not there yet.  

So which is it? It matters, because Microsoft and OpenAI are furiously negotiating over whether the latter’s models can be more broadly available, whether OpenAI can go public, and how much Microsoft’s stake in the company will be. 

How Do We Measure AI and AGI? 

To know if we’ve achieved AGI, we have to be able to measure it. 

Several organizations have proposed tests. Lab42, headquartered in Davos, Switzerland, offers a battery of reasoning tests to measure just how “intelligent” an AI can be. Here’s a sample problem from their website:  

Google’s DeepMind approaches the problem differently. It suggests a 2×2 matrix classifying “Levels of AGI,” in effect describing an AI as either narrow (can only handle certain types of problems) vs general (can handle a wide range of problems) on the one axis; and performance, that is, the extent to which an AI can mimic humans, on the other. A Level-0 application could be (for example) a calculator, with no AI; Level-5, somewhat terrifyingly, is termed “superhuman,” exceeding human capabilities. (Luckily, we haven’t achieved that yet.)  

And there are lots of other possibilities. We could simply give it IQ tests, or current events tests, or empathy-emulation tests, or all of these and more.  

Ultimately, I suspect, it will come down to a value judgment based partly on how well a given AI performs on quantitative tests – and how human it “feels.”  

Which will make that pesky clause in the Microsoft-OpenAI contract difficult to resolve.  

Efficiency and Innovation 

Today, organizations spend large sums of money on fundamentally two types of initiatives: innovation and operational efficiency. (It’s true when you think about it.)  

Every AI use case I’ve seen is about operational efficiency. AI can make your employees better writers, better coders, in general more productive – in many cases much more productive (full disclosure: I used AI to research this article – but not to write it.) 

I’ve yet to see an AI that’s truly innovative. It’s difficult to imagine that AI could, independent of any human assistance, come up with a revolutionary product like the iPhone, for example. Sure, it could make an existing product better, improve production times, and so on – but true innovation?  

To me that’s true “intelligence” and that’s our job, and always will be. 

Progress, Determinism and Probability 

Lastly: many have commented that our march to AGI has slowed, that progress in AI has plateaued. True, training and inferencing have become more efficient (see previous paragraph): Nvidia’s chips have gotten faster and denser, data centers are springing up everywhere (remember, capex=strategy) And OpenAI and others are testing new techniques like reinforcement post-training fine-tuning (basically you tell the LLM the right answer and it adjusts its weights accordingly).  

Thus with continuing incremental improvements and the economics of scale, on the whole things arguably will gradually get better and perhaps cheaper (we’ll see).  

But let’s remember: LLMs today are fundamentally probabilistic: they use advanced statistical algorithms to infer what the next best word or next best sentence in a response to a prompt might be. (Is post-training RL “cheating?” You be the judge.) That’s very different from how traditional, deterministic software applications work, in which you are guaranteed the same output for the same input. 

In my opinion, what’s missing in current LLMs is a solid grounding in common-sense knowledge. “A dog is a mammal” – everybody instinctively knows that. Not an LLM, though – it reasons over its training base to determine that statement’s probability. It’s usually going to get it right – but it’s not guaranteed.  

Therefore I suspect that the next major leap in AI will occur when LLMs can refer to a systematized ontology of indisputable facts to reduce probabilistic errors and hallucinations.  

Net: it may be a while before an AI can truly pass as human. 

Microsoft and OpenAI

Let us know!  

Yes, we’ve digressed (but wasn’t it fun?).  As I write this, Microsoft and OpenAI are trying to resolve their differences – who will win? Does it matter? What do you think?  

I’m at bbriggs@directionsonmicrosoft.com.  

Microsoft is increasing its campaign to give Windows 10 customers a safe exit strategy before it cuts off support (and specifically security updates) for that OS version in October 2025. Microsoft is doing this in two ways: By trying to get more individuals to sign up for Windows 10 Extended Security Updates (ESUs) and by going public with more details of how it plans to make Windows 11 more secure — and hopefully more compelling. 

With less than four months to go before the end of Windows 10 support on Oct. 14, 2025, Windows 11 adoption remains lackluster. Most independent estimates claim as many as half or more of all Windows PCs are still on Windows 10. Some of these users just don’t like Windows 11; others cannot install it because their older PCs do not meet the hardware requirements. And unsupported Windows 10 PCs will make the overall Windows ecosystem less secure by making endpoints vulnerable to attacks.

Individuals Get More Windows 10 ESU Options 

Earlier this week, Microsoft announced it would give users a couple of additional ways to get one year of extended support for Windows 10, enabling them to keep getting important security updates beyond Oct. 2025. 

Enterprises already can enroll in the Windows 10 ESU program via a Microsoft Volume Licensing program. They also will get the option to purchase ESUs from Cloud Service Providers starting Sept. 1.  

As previously announced, Microsoft is charging commercial customers USD $61 per device for Windows 10 ESUs for the first year, with the price doubling in Years 2 and 3. Microsoft is charging educational institutions USD$1 per device for Windows 10 ESUs for Year 1, with the price doubling in Years 2 and 3. The only way Windows 10 commercial customers can get Windows 10 ESUs for no additional charge for three years beyond the Oct. 2025 cut-off is to subscribe to Windows 365 and use it from a  Windows 10 device locally or run a Windows 10 instance in an Azure Virtual Desktop. 

Individuals/consumers running Windows 10 have a couple of new options to get ESUs. In addition to the already announced $30 USD single-year ESU option, Microsoft will allow individuals to redeem 1,000 Microsoft Rewards points to pay for the one-year ESU for Windows 10 or use Windows Backup to sync their settings to the cloud and get the one-year ESU for free. Starting in July, customers will see these new ESU options via an enrollment wizard that will be available via notifications and in Settings on their personal Windows 10 PCs. 

Microsoft announced in May that it was extending the period during which it would provide security updates for Microsoft 365 apps running on personal and commercial Windows 10 PCs through Oct. 10, 2028. They also will get feature updates for Microsoft 365 apps through August 2026, Microsoft officials said. 

Making Windows 11 More Resilient 

Microsoft also provided more specifics this week around its Windows Resiliency Initiative, which it announced last year. One of the biggest changes, which Microsoft will begin testing in private preview among select security vendors in July, is moving antivirus and endpoint detection and response (EDR) apps out of the Windows kernel. Microsoft has been working with a variety of security vendors to head off update meltdowns like the one from CrowdStrike that hit millions of Windows machines worldwide last year. It’s not clear when these changes will make their way to Windows 11 customers 

Microsoft also is planning to make available later this summer a new Quick Machine Recovery feature that is meant to help organizations quickly restore machines that cannot boot. Microsoft said the QMR feature will be generally available to all Windows 11 24H2 devices starting this summer. And as part of that summer update, Microsoft also will be replacing the Blue Screen of Death with a new Black Screen of Death that will provide clearer information on the screen about issues requiring reboots. 

In addition, Microsoft is rolling out “in the coming weeks” a feature called Connected Cache that is meant to help organizations improve bandwidth when upgrading to Windows 11, handling Windows Autopilot device provisioning, performing Intune application installations and incorporating Autopatch monthly updates. 

Microsoft also announced recently plans to clean up legacy drivers that Windows Update pushes to users’ devices when installing updates. Microsoft is making the change to reduce security and compatibility risks, officials said. 

Related Resources 

More Options for Windows 10 ESUs 

How to Enable Extended Security Updates for Windows 10

Update on the Windows Resiliency Initiative 

Windows Resiliency Initiative Will Affect Windows and Anti-Malware (Directions members only) 

Microsoft is adding a new Sovereign Private Cloud option to its set of cloud services available in Europe. It also is expanding its existing Cloud for Sovereignty service with new features for European customers and renaming that service to “Sovereign Public Cloud.” 

Microsoft announced its new Sovereign cloud strategy for European customers on June 16. Earlier this month, rival AWS announced plans to launch later in 2025 a new European Sovereign Cloud company, designed to be locally controlled in the European Union, staffed by EU citizens and subject to EU laws. Last month, Google updated its own sovereign cloud services and added a disconnected, air-gapped option for customers with strict data security requirements, as well as a new Google Cloud Dedicated option for local and regional partner deployments. 

Calls for European countries and customers to reduce their reliance on the biggest U.S. cloud providers due to fears of political and trade havoc have definitely hit a nerve with Microsoft and other leading cloud companies.  Microsoft’s new sovereignty announcements come on the heels of it announcing five new digital commitments meant to reassure European customers that Microsoft has their back, in spite of the chaos caused by the current U.S. government regime with tariffs, regulation, threats over data access and more.  

Expanding the Public Sovereign Cloud Footprint 

Up until now, Microsoft’s Sovereign cloud offering, launched in 2022, initially was built to host Microsoft 365, Dynamics 365 and Azure. Microsoft positioned it as an offering for highly regulated customers and government agencies throughout the world in select markets worldwide. At the time, Microsoft defined “sovereignty” fairly loosely; It was a term describing where data resided, but also applied to security, compliance, and policy requirements that were particular to various countries’ governments. 

The Sovereign Public Cloud, currently in preview, is slated to be generally available in all European Azure regions in 2025. Sovereign Public Cloud is getting a new Data Guardian feature meant to ensure EU customers’ data is stored and processed exclusively in Europe and controlled exclusively by Microsoft employees based in Europe. The Sovereign Public Cloud also is now supporting customers and/or their partners who want to control the key management for the encryption of their data. Additionally, the Sovereign Public Cloud is getting a new service for administering the sovereignty functions called Regulated Environment Management. 

“Sovereign Public Cloud ensures customer data stays in Europe, under European Law, with operations and access controlled by European personnel, and encryption is under full control of customers. This is enabled for all customer workloads running in our European datacenter regions requiring no migration,” Microsoft officials said. 

Sovereign Private Cloud features Azure Local, Microsoft 365 Local 

Microsoft’s new Sovereign Private Cloud offering is for those who want to run their apps either on site, in partner datacenters or elsewhere in their own country. Microsoft says it will support hybrid or air-gapped (isolated) operations, and is built on top of Azure Local, the product formerly known as Azure Stack HCI. 

The Private Cloud option also supports “Microsoft 365 Local,” which brings Office server software like Exchange Server and SharePoint Server to Azure Local running in a customer’s own datacenters or sovereign cloud environments. Microsoft says Private Sovereign Cloud is for governments, critical industries and regulated sectors needing the highest standards of data residency, operational autonomy and disconnected access. 

As part of its June 16 announcements, Microsoft also gave the existing set of Sovereign Cloud offerings run by select partners a new name. These are now called “National Partner Clouds.” Partners working with Microsoft on these clouds include Accenture, Crayon, Capgemini, Dell Technologies, IBM, NTT Data, Orange, Telefonica, Vodafone and more. 

Directions on Microsoft asked Microsoft whether it plans to make these new Sovereign Public Cloud features and the new Private Sovereign Cloud offering available beyond the EU. We asked for any pricing and licensing details that the company can share. We also asked if Microsoft planned to make Microsoft 365 Local available outside of its Private Sovereign Cloud service. So far, no word back on any of these questions. 

Update (June 17): A Microsoft spokesperson replied with answers to all three of our questions.

Regarding pricing and licensing information: “Microsoft’s Sovereign Public Cloud and Sovereign Private Cloud solutions are currently in preview and set to be generally available in all European cloud regions later this year. We will share more details on costs, if applicable, as offerings become available.”

Regarding making the updated Sovereign Cloud offerings available beyond Europe: “We are starting with 15 EU/EFTA countries and will add additional countries and regions, aligned to customers’ needs.”

Regarding making Microsoft 365 Local available outside of the Sovereign Private Cloud: “Microsoft 365 Local is a deployment and management framework that enables customers to run Microsoft productivity workloads like Exchange Server and SharePoint Server on Azure Local infrastructure. We are not announcing licensing changes for those products as part of this approach.”

Microsoft is warning Microsoft 365 enterprise customers of some changes coming in July 2025 that could affect how and how often organizations will get Office app updates on Windows desktops. Organizations that have been juggling updates via the Semi-Annual Enterprise Channel in a way that allowed them to only update their apps once a year are going to be forced to update twice annually, thanks to the coming new policies. 

Microsoft published information on pending update changes to the Microsoft 365 Message Center (MC1087098) on June 2. Administrators need to take action quickly to ensure desktops continue to get regular feature and security updates. 

Microsoft officials said they are making the changes in order to “streamline updates and align with customer needs.” In other words: We are rolling out new features, especially AI/agentic ones, at a fast clip and we want customers to get access (and pay us, if possible) as soon as possible. 

Currently, Microsoft offers three primary update channels for M365 Enterprise App customers: The Current Channel; Monthly Enterprise Channel; Semi-Annual Enterprise Channel. (It also has a Semi-Annual Enterprise Preview Channel and Semi-Annual Enterprise Extended Channel.) Microsoft uses these channels to provide both feature updates and security and non-security updates for Microsoft 365 desktop apps. 

Up until now, those who opt for Current Channel get new Microsoft 365 Apps features as soon as they’re ready, which means typically at least once a month. Monthly Enterprise channel gets new or updated app features on the second Tuesday each month. Those in the Semi-Annual Enterprise Channel also get feature updates on the second Tuesday, but only in January and July. Semi-Annual Channel Preview is for customers who want to preview new features coming to the Semi-Annual Channel ahead of deploying them. The Semi-Annual Extended Channel is for updates delivered to subsequent releases of previous forks. 

What’s Happening When 

Beginning in July 8, 2025, Microsoft will no longer provide updates via the Semi-Annual Enterprise Channel Preview. Officials said organizations should move devices on this channel to another supported channel if they want to keep getting early access to new features and security updates. And as of March 10, 2026, it is dropping support for Semi-Annual Enterprise Channel Extended. 

Microsoft is making other channel changes starting in July 2025. The Monthly Enterprise Channel will get rollback support for two months, up from one. The Semi-Annual Enterprise Channel will be supported for fewer months — from 14 months to 8 months. (The eight months includes 6 months of regular support plus the 2-month rollback period.)  

The Semi-Annual Enterprise Channel will be meant for unattended devices, going forward, according to Microsoft. (Up until now, Semi-Annual Enterprise Channel was for “non-interactive devices and those running specialized or business critical workloads that require extensive testing” before new features are implemented.) 

“Bottom line: You have to deploy new Office versions twice a year if you are on the Semi-Annual Enterprise channel – up from once,” said Directions on Microsoft analyst Rob Helm. 

Currently, via the Semi-Annual channel, Microsoft supports new Office app releases for 14 months. That means that enterprises can deploy new versions of Microsoft 365 Enterprise apps in January, ignore the July updates and then deploy the newest versions the following January. The supported roll-back window meant if there were any issues with the new Office apps, companies could revert to a version that was more than a year old until the problems were fixed or a newer version of the apps was available.  

Organizations can change the Microsoft 365 Apps update channel for their devices using Group Policy, the Office Deployment Tool (ODT), Configuration Manager, Intune Administrative Templates, and/or the Microsoft 365 Apps Admin Center. 

Related Resources 

Microsoft 365: Retiring and modernizing the Semi-Annual Enterprise Channel for Windows desktop apps  

Updates Coming to Microsoft 365 Apps Channels 

From 2020: Microsoft starts pushing enterprise to pick up the Office app update pace 

How to change Office update channels 

Microsoft has once again “clarified” rules that change how Dynamics 365 user license compliance is measured but could also simplify how to determine what user licenses are required. Many Dynamics 365 Finance + Operations customers should spend time now cleaning up security roles assigned to their users, before Microsoft potentially claims they are out of compliance. (Note: this change does not impact Dynamics 365 CRM customers.) 

What Happened?

On May 3^rd, 2025, I was going through the latest Dynamics 365 Licensing Guide to look for changes, when I noticed the addition of a sentence in Appendix E on page 56:

“Licensing requirements for these applications are determined by the role-based security assigned to each user.” 

“These applications” refers to the Dynamics 365 Finance and Operations applications: Finance, Supply Chain Management, Commerce, Project Operations, and Human Resources. 

The significance of the sentence is that the user licensing requirements for Dynamics 365 Finance and Operations applications have shifted from “what a user actually did” to “what a user could potentially do.” 

Previously, organizations were on the hook for a Dynamics 365 license based on the application users accessed, the routines they ran, and the type of data they updated. Microsoft provided some reports that gave guidance on what license users “might” require, but the ultimate test of license compliance was based on activities users performed. This could only be determined by viewing audit logs, which had their own limitations, or by evaluating the custom screens or custom applications users accessed, because these custom screens and applications programmatically limited what users could do. (In Nov. 2024, I wrote extensively about how to tackle user license compliance for our members in “Dynamics 365 User License Compliance: A Struggle that Benefits Microsoft.”) 

However—and this is a big however—going forward, the new sentence “clarifies” that what a user actually does is no longer relevant, because organizations are now on the hook for a Dynamics 365 license or licenses that cover security roles assigned to users. 

Why is this a Concern?

I’m concerned for companies who have enjoyed Microsoft’s relaxed approach to user license compliance and have routinely handed out higher security roles than needed, which now exposes the organization to new (sorry, “clarified”) license requirements. 

The licensing change won’t severely impact everyone, especially not those customers who were diligent with assigning security roles, but I do think it will be a problem in the following scenarios: 

Security roles applied for convenience. Many admins took the easy approach and assigned users with as many security roles as needed to get the user working and meeting any potential request, reducing the need for the user to come back and ask for more rights later. Is this a best practice? No. But we all know that it happens. 

Customizations using default security roles. This is the biggest issue. Most customers customize Dynamics 365 Finance + Operations in some way. They modify Dynamics 365 screens to remove certain fields or add access to other data, and some customers build or purchase custom applications that augment the Dynamics 365 service and access data directly. In these situations, an admin must still assign users with a security role, and those security roles (especially the default ones) often contain rights beyond what a user needs for the custom scenario. This is generally acceptable, because the custom screen or application limits what a user can do, so giving them excessive rights is a low risk. 

However, with the new sentence in the Licensing Guide, all users are required to have a license based on the security roles they are assigned, whether or not they actually perform activities allowed by those security roles. 

What Should Customers Do? 

Don’t panic. Although the change is in place as of May 2025, there is no automatic compliance check that will lock users out if they don’t have the right licensed assigned. What you should do is start planning for your next audit or renewal, so I recommend the following steps. 

1. Talk with your Dynamics 365 security admins and find out how big of an issue this is. They will have the best sense of the magnitude and risk for your organization. 

2. Adopt a “least access” approach to security roles. In other words, do the right thing and only assign users the security roles and features they need, rather than being overly generous. In many cases, this will involve building custom security roles that only grant rights to the features and data users actually need. Custom security roles are a common approach and nothing to be feared. 

3. Clean up existing security role assignments. Microsoft is previewing new license compliance reports that compare the security roles a user has (and therefore the license required) with the licenses users are assigned. The reports can highlight when users are under licensed, but I suggest using it as a starting point to determine if users have excessive security roles or could benefit from a custom security role that has a lower licensing requirement. 

4. Talk with developers who customize the Dynamics 365 service and determine if they used default security roles or custom security roles to provide access to their custom solution. Typically, custom solutions benefit from custom security roles that grants the least number of rights and therefore the lowest possible license requirement. 

Do you really read the Licensing Guide each Month? 

It’s been asked this many times and yes, I read through the new Licensing Guide every month, looking for changes that impact customers.  

I also keep a copy of most of them, in case I need to look back through time. I’m not sure anyone needs a “Dynamics 365 (On-premises) Enterprise Edition – Dec. 2016” licensing guide, but I have one. 

In a similar situation, last April, I wrote about how Microsoft “clarified” licensing rules about Dynamics 365 and Power Apps last April as well in “Dynamics 365 rights in Power Apps Subscriptions: Changes on the horizon.” 

Related Resources

Directions kit “Dynamics 365.” (Directions members only)

Licensing Reference Set for Dynamics 365 (Directions members only)

Microsoft has announced the planned retirement of SharePoint Alerts for SharePoint Online by July 2026.  SharePoint Alerts have been around since the dawn of time so if you have used SharePoint for a while in your organization, there’s a good chance you and your users will be affected as early as September 2025.  

If your organization is new to SharePoint, don’t get started now with Alerts. Evaluate using SharePoint Rules or Power Automate which are now the preferred methods of sending alerts (aka notifications). 

SharePoint Alerts are email or text notifications users can set up to track activity in the SharePoint sites they use.  Alerts apply to lists and document libraries and can be an extremely helpful tool when used wisely.  When users configure an alert, they can select the type of change that will trigger them, such as when anything changes, when someone else changes an item, or when someone else changes an item created or modified by the user.  To manage notification overload, users can also specify how often to receive alerts: immediately, daily or monthly.  Since this is a user-managed tool, you can imagine how many of these alerts may exist in your environment. 

SharePoint will turn 25 years old next year.  There’s a lot of legacy technology in the platform and the alert mechanism is among them.  It’s being replaced by a choice between SharePoint Rules and Power Automate, both using more modern technologies.  That doesn’t mean there won’t be pain.  Users are accustomed to the current alerts and must be prepared for the switch.  It’ll be a new experience which means organizations will have to educate and train them on which approach will be used (or both) and how to adjust their ways of working. 

What’s Microsoft’s Plan (and Yours)?

While Microsoft announced the end of SharePoint Alerts will be July 2026, it will begin turning off new alert creation as early as July 2025 for customers new to SharePoint.  Existing customers may start seeing new alert creation turned off starting in September of 2025.   

Start planning for the transition and exploring the modern approaches of SharePoint Rules and Power Automate.  To help with planning, you can use the Microsoft 365 Assessment tool to identify Alert usage and begin to understand the size of the task at hand. Users will begin seeing warning banners on various configuration pages in SharePoint so it’s important that you have a plan to communicate it out to your organization before panic ensues. 

Evaluate your options between SharePoint Rules and Power Automate.  You will have to determine when to use what and educate your users on the approach(es).  SharePoint Rules are native to SharePoint, but Power Automate is licensed and may incur additional costs.    

Communicate your plan to your users. And communicate it again. 

The Wrap Up

At this time, there is no indication from Microsoft of a migration approach from alerts to either SharePoint Rules or Power Automate.  This also only appears to apply to SharePoint Online for now.  However, since Microsoft removes other legacy technology from SharePoint Server, it may be coming to you in the future.   

Is this important to you?  Let me know what else you want to hear about by emailing me at dberry@directionsonmicrosoft.com.  

Microsoft’s annual Build developer conference is happening this week in Seattle. No big surprise: All things AI and, specifically, agents, are at the top of the announcement list. Surprise: There was quite a bit of news for those still developing on Windows, too. 

In no particular order, here are my top 10 takeaways from Day One of Build: 

1. Multi-agent is the new hot.  Organizations will soon be able to build multi-agent systems in Copilot Studio, where agents delegate tasks to one another. The agents can be built with the Microsoft 365 Agent Builder, Azure AI Agents Service, and Azure Fabric. The Copilot Studio capability to build multi-agent systems is now in preview. 

2. NLWeb is a way to turn websites into agentic apps. Natural Language Web (NLWeb) is a Microsoft-developed project for simplifying the natural language interface for websites. (It’s not clear what, if any, connection that OpenAI has with NLWeb. Other current sponsors of NLWeb include O’Reilly Media, Snowflake, Shopify, Chicago Public Library.) Microsoft says NLWeb was conceived and developed by R.V. Guha, former Google Tech Fellow who recently became a Microsoft CVP and Tech Fellow. Guha is the creator of RSS, RDF and Schema.org. 

3. GitHub Copilot is moving from pair programmer to peer programmer. GitHub Coding Agent (“Project Padawan”) is now generally available to all Copilot Enterprise and Copilot Pro+ customers. The agent is best suited to low-to-medium complexity tasks in well-tested codebases and can handle things like adding features and fixing bugs to extending tests, refactoring code, and improving documentation, Microsoft says.  

4. Copilot Studio is getting more and more “pro” dev features. As was evident at Ignite last fall, Microsoft increasingly is positioning Copilot Studio, its low-code/no-code development platform, as a tool for professional developers who need more powerful tools to build more complex agents. Microsoft is working on exposing a set of M365 Copilot APIs for developers. (So far, only the retrieval API is available in preview.) Microsoft is working to close the gap between Foundry and Copilot Studio by enabling a Bring Your Own Models from Azure Foundry capability, which is currently in preview. 

5. The MCP protocol is becoming increasingly key to building agents. Microsoft has joined the Model Context Protocol (MCP) Steering Committee is integrating MCP support across GitHub, Copilot Studio, Dynamics 365, Azure, Azure AI Foundry, Semantic Kernel, Foundry Agents and Windows 11. “Think of MCP as a universal USB-C connector for AI,” linking apps, agents and tooling in a standardized way. 

Windows: The Best AI Dev Platform?

6. The Windows 11 Copilot Runtime has been renamed “Windows AI Foundry.” AI Foundry is now Microsoft’s brand for the local-platform versions of Azure AI Foundry and are built for model selection, optimizing, fine-tuning and deployment across client and cloud. Microsoft officials say that Windows ML is the built-in inferencing runtime on Windows. Microsoft also announced a local version of AI Foundry for the Mac. Both of these local Foundry offerings are generally available. 

7. Microsoft is open sourcing the Windows Subsystem for Linux (WSL). They’ve already open sourced the “Mariner” Linux underpinning and the WSL-g UI layer. Now it also is open-sourcing most of WSL itself for “reasons.” Maybe the community will do something with direct access to WSL APIs? Microsoft also announced plans to open source the GitHub Copilot Chat Extension for VSCode “in the coming weeks.” 

8. Microsoft is expanding Entra, Defender for Cloud and Purview by embedding them directly into Azure AI Foundry and Copilot Studio to help organizations secure AI apps and agents across the entire development lifecycle. Entra Agent ID, a capability meant to manage built using Microsoft tools, as well as across select third parties like ServiceNow and Workday, is now in preview. 

9. SQL Server 2025 is now in public preview. The coming release of SQL Server is now available to testers. Microsoft is touting the integration of AI directly into the database engine “enabling more intelligent search.” It also has built-in vector search capabilities. 

10. OpenAI CEO Sam Altman and Special Government Employee and xAI Founder Elon Musk both appeared in Nadella’s keynote — via video. Musk talked up Microsoft hosting Grok 3 and Grok 3 Mini in Azure Foundry (like it already does with thousands of third-party models, including DeepSeek). Altman talked about OpenAI’s announcement of its Codex programming assistant (while making sure to show continued excitement about the great Microsoft partnership OpenAI enjoys).

To read about all of the Microsoft Build 2025 announcements, check out the Book of News. Fun Fact: There are more than 300 mentions of “agent” in Microsoft’s Build news. And around 30 mentions of “Copilot.” It’s hard to keep up, especially when it comes to AI branding at Microsoft….

Microsoft has announced it will provide security updates for Microsoft 365 apps running on Windows 10 until Oct. 2028, rather than until Oct. 2025, as planned. 

The company published an article on Microsoft Learn announcing the change on May 8, 2025. (Thanks to Neowin for the original link.)

Microsoft 365 apps would have continued to work on Windows 10 after Microsoft support ends for the operating system on Oct. 14, 2025, the company acknowledged, but they would have run unsupported. To try to deter customers from running these apps on Windows 10 after support ends, Microsoft officials have said running these apps on an unsupported OS could potentially cause “performance and reliability issues.” 

Microsoft is now planning to provide security updates to the Microsoft 365 Apps on Windows 10 until Oct. 10, 2028 “to help maintain security while you transition to Windows 11,” the Learn article notes. The new security update also applies to the subscription versions of the Project and Visio desktop apps, official said. 

The new security update extension does little, if anything, regarding Microsoft-provided support for M365 apps running Windows 10 after Oct. 14, 2025, the Microsoft Learn article says.  

If a customer with a valid Microsoft 365 subscription has issues with their apps, they will continue to be able to open support cases, but Microsoft support will provide only troubleshooting assistance. “Technical workarounds might be limited or unavailable,” the article says. Whether or not customers on Windows 10 have ESUs, they do not have the option to log a Microsoft 365 apps bug or request other product updates. 

Will Windows 10 Get a Security Reprieve, Too? 

Directions on Microsoft has asked Microsoft if there will be a similar three-year pushback on the date when security updates will no longer be provided for Windows 10. No word back so far. 

Update (May 13): A Microsoft spokesperson declined to answer the question we posed and instead provided a link to an April 2024 Microsoft blog post on when to use Windows 10 ESUs.

We’d point out that Microsoft provided an additional three years of security updates for Microsoft 365 apps on Windows 7 when Windows 7 was approaching its end-of-support date.

Given that Microsoft has announced it will be selling Windows 10 Extended Security Updates (ESUs) for three years and to consumers for one year, we’re thinking a delay on the end of support for Windows 10 seems unlikely. Microsoft is charging commercial customers USD $61 per device for Windows 10 ESUs for the first year, with the price doubling Years 2 and 3.  

However, if there are a substantial number of Windows 10 business and personal PCs still running after Oct. 10 this year, the security risk caused by these systems will be substantial. Various market watchers have estimated that fewer than 50 percent of Windows PCs are running Windows 11. Microsoft is continuing to advise customers to move off Windows 10, either by upgrading to Windows 11 on eligible PCs, licensing Windows 365, or by buying new PCs that support Windows 11. 

Related Resources

Windows 10 end of support and Microsoft 365 Apps 

Meteor 2025: The Next Wave of Windows and Office Migrations (Directions members only) 

Directions Licensing Reference: Extended Security Updates for Windows 10 (Directions members only)