AI agent
Also called LLM agent, agentic AI.
Software where a language model decides which steps to take and which tools to call, in a loop, to reach a goal.
Last updated
What it is.
An agent gives an LLM a goal, a set of tools and a loop. At each step the model looks at what it has so far, picks a tool to call, such as search, a database query or sending a message, reads the result and decides what to do next, until it finishes or hits a limit.
The difference from a fixed workflow is who decides the path. In workflow automation, a person designs the steps in advance. In an agent, the model chooses them at run time. Many real systems are a mix: a fixed workflow with one agent step where judgment is needed.
Why it matters
Agents can handle tasks where the steps vary from case to case, such as researching a request, triaging a ticket or reconciling records across systems. That flexibility costs predictability: more model calls per task, more ways to fail and more need for guardrails, logs and human review on actions that matter.
Common mistakes.
- 01
Using an agent where a fixed workflow would do. If the steps are always the same, write them down.
- 02
Giving the agent broad permissions. Scope each tool to the least access it needs.
- 03
No step limit or budget. A confused agent can loop and run up cost.
Work it out with a tool.
Related terms.
- Tool calling (function calling)A way for a language model to ask your application to run a defined function and give it the result.
- Guardrails (AI)Checks placed around an AI model that block, fix or flag unsafe, off-topic or incorrect inputs and outputs.
- Human-in-the-loop (HITL)A design where a person reviews, corrects or approves what an AI system produces before it takes effect.
- LLM evals (evaluations)Repeatable tests that score how well a model, prompt or AI feature does a specific task.
