AI Agent
In one sentence
An AI that can read files, call external tools, and complete multi-step tasks on its own, unlike a standard chat AI that only replies with text.
A chat AI stops once it answers a question. An AI agent takes a task, breaks it down into steps—such as searching for references, reading files, drafting copy, and saving the file to a specific folder—and adjusts its approach if it runs into an issue along the way.
What an AI agent can accomplish depends on the tools it can access, the reference materials it can read, and whether it has clear operational instructions (SKILLs).
Example
An employee asks to "summarize last week's sales reports into a one-page brief and email it to the department manager." The AI agent reads the report files on its own, compiles the data, generates the document, and sends it out.
Related terms
- The Three Pillars of AI Adoption
- Starbox's framework for putting AI to work: AI Agents do the work, SKILLs standardize how each task is handled, and Knowledge Bases store your company's data and institutional experience.
- SKILL
- A standardized operating procedure written for an AI agent that defines how to run a specific task, the sequence of steps, which materials to use, and the final delivery format.
- Tool Calling (Function Calling)
- As an LLM generates a response, it calls external tools as needed—such as search, calculations, database queries, or messaging—and continues working based on the results.
- Workflow
- Break a task into fixed steps and assign them in sequence to AI or humans, with predefined inputs and outputs for each step.
Further reading
AI News
Who Owns an AI Agent After the Employee Leaves?
When employees build custom AI agents, turnover turns personal assets into operational risks. Ownership handover is now a core enterprise requirement.
Oct 12, 2026 · 3 min read
AI News
Your AI Agent Has Tool Access. But Which Records Can It Actually See?
Granting an AI agent tool access does not mean it should see every record. Object-level access control sets boundaries before data reaches the model context.
Sep 27, 2026 · 4 min read
AI News
Do Not Govern Every AI Agent at Once: Build a Risk Priority Queue
An AI agent inventory is not enough. When resources are limited, enterprises need a risk queue to decide which agents to review and control first.
Sep 25, 2026 · 4 min read
