AI agent: An AI agent is a system that uses an AI model to pursue a goal on its own by planning steps, using tools such as web browsers, files, and software APIs, and adjusting based on results, rather than answering a single prompt.
A chat assistant answers and stops. An agent keeps going. Give it a goal ('find three vendors that meet these requirements and draft an outreach email to each'), and it decides what to search, reads the results, opens pages, takes notes, drafts the emails, and reports back, looping until it is done or stuck.
What makes this possible is tool use. The model can call a search engine, run code, read and write files, query a database, or operate a browser. Coding agents such as Claude Code and GitHub Copilot's agent mode write and test software this way, and general-purpose agents in ChatGPT, Claude, and Microsoft Copilot can complete multi-step research and office tasks.
Agents multiply both the value and the risk of a mistake. A wrong assumption early in a twenty-step task compounds, and an agent with access to email or payment systems can act on it. Start with read-only tasks, review before anything is sent or changed, and give agents the narrowest access that gets the job done.
Example at work
An operations analyst asks an agent to reconcile this month's carrier invoices against the shipment log in a spreadsheet, flag discrepancies over $50, and draft dispute emails for each. The agent works through the files and produces a discrepancy list and draft emails. The analyst reviews the list and sends the emails herself.
Why it matters
Agents are where AI shifts from helping you do a task to doing the task while you supervise. The professionals who benefit most will be the ones who can specify a goal clearly, set sensible limits, and check the work, the same skills as managing a capable new hire.
Related terms
- Agentic AIAgentic AI describes AI systems designed to act with autonomy: they take a goal, break it into steps, use tools, and carry the work through with limited human direction. It is the broader approach; an AI agent is a specific system built that way.
- Tool useTool use (also called function calling) is the ability of an AI model to request that an outside tool be run, such as a web search, a calculator, a database query, or a calendar lookup, and then use the result in its answer.
- Workflow automationWorkflow automation connects apps and steps so that a trigger, such as a new form response or an incoming email, automatically runs a sequence of actions, increasingly with an AI step that reads, decides, or writes along the way.
- Human-in-the-loopHuman-in-the-loop is a design approach in which a person reviews, approves, or corrects an AI system's output at defined points before it takes effect, combining the speed of automation with human judgment and accountability.
- Model Context Protocol (MCP)The Model Context Protocol (MCP) is an open standard, introduced by Anthropic in late 2024, that lets AI assistants connect to external tools, data sources, and services through a common interface instead of custom one-off integrations.