What Is an Agent: Executive Overview

• The word “agent” now refers to a program that uses AI to reason over problems, plan, and execute solutions.
• Agents can often perform tasks, especially tedious ones, better and faster than humans.
• However, agents can bring security, cost, governance, compliance, and other risks.

Agents have quickly become a ubiquitous construct in IT offerings, proliferating AI in numerous platforms. Conceptually understanding the design, capabilities, and risks of agents has become critical across industries and disciplines.

What Is an Agent?

Agents are programs that leverage AI, whether working interactively with a human user in natural language or running in the background without ongoing user direction. Agents can perform prescribed routines, similar to traditional workflows, or they can act autonomously with AI-driven reasoning to make decisions on how and when to perform tasks.

Agents primarily differ from traditional programs and services due to their supplemental AI, which boosts their logic and reasoning capabilities. Whereas prior programs could only respond to developer-specified inputs and conditions, agents use AI to decide what to do and how to do it without ongoing human intervention. The ability to reason and act without a developer describing every possible situation and reaction procedure gives agents a significant advantage for speeding existing solutions and enabling new ones.