# AI agent

> An AI agent is a language model that chooses and calls tools in a loop, reading each result, until it finishes a task or hands off to a person.

- Page: https://methodcrew.com/glossary/ai-agent
- Last updated: 2026-10-28
- Publisher: [The Method Crew](https://methodcrew.com/)
- Also known as: LLM agent, autonomous agent

## How it works

An agent gets a goal, a set of tools and rules for when to stop. Tools are functions it is allowed to call: search the CRM, read an inbox, create a ticket, run a database query. On each turn the model looks at the goal and everything it has seen so far, picks a tool, gets the result back and decides what to do next. The loop ends when the task is done, a step limit is hit or a rule says a person must approve.

Anthropic's engineering guide, [Building effective agents](https://www.anthropic.com/engineering/building-effective-agents) (December 2024), draws a useful line. Workflows run model calls along code paths you define in advance. Agents choose their own path. Most production systems are mostly workflow, with an agent handling one or two steps that need judgment.

## How businesses use it

Good first agents have a narrow job and a clear finish line:

- Qualifying inbound leads: read the form, look up the company, score it against your criteria, then book a call or route it
- Triaging support email: classify it, pull the order history, draft a reply for a person to send
- Preparing account research before sales calls
- Reconciling records across 2 systems and flagging what does not match

A first [custom AI agent](https://methodcrew.com/services/custom-ai-agents) build should be one of these, on your own accounts, with model usage billed at cost. The [inbound lead qualification blueprint](https://methodcrew.com/blueprints/inbound-lead-qualification-agent) shows a narrow agent built as a workflow, from intake to a drafted reply.

## Where it goes wrong

Agents cost more than fixed workflows and fail in stranger ways. Each extra tool and each extra step is another chance for a wrong turn, and errors compound across a long loop. Typical failures: the agent calls the wrong tool, passes a made-up order number as an argument, loops without finishing, or takes an action nobody approved.

If the steps are the same every time, build a workflow in n8n or Make instead. It is cheaper to run and easier to debug, and our comparison of [agents, automation and chatbots](https://methodcrew.com/insights/ai-agents-vs-automation-vs-chatbots) covers that decision. When you do need an agent, give it the fewest tools that do the job, log every call and put [human-in-the-loop](https://methodcrew.com/glossary/human-in-the-loop) approval in front of anything that emails a customer or moves money.

## Example (hypothetical)

Hypothetical: a 40-person freight broker gets 200 quote requests a week by email. An agent reads each request, checks lane rates in the transport management system and drafts a quote into a review queue, where a coordinator approves or edits it before anything is sent.

## Related terms

- [Agentic AI](https://methodcrew.com/glossary/agentic-ai): Agentic AI is the umbrella term for systems where language models plan steps and take actions through tools, with less step-by-step direction from people.
- [Function calling (tool use)](https://methodcrew.com/glossary/function-calling): Function calling lets a language model ask your software to run a named function, such as a CRM lookup, by returning the function name and arguments as JSON.
- [Human-in-the-loop (HITL)](https://methodcrew.com/glossary/human-in-the-loop): Human-in-the-loop is a workflow design where a person reviews, corrects or approves an AI system's output at set points before it reaches a customer or a system of record.
- [Model Context Protocol (MCP)](https://methodcrew.com/glossary/model-context-protocol): The Model Context Protocol is an open standard for connecting AI applications to outside tools and data, so a single integration works across compatible assistants and agents.
- [AI guardrails](https://methodcrew.com/glossary/ai-guardrails): AI guardrails are checks placed around a language model that block, rewrite or escalate inputs and outputs that break your rules on safety, accuracy, privacy or scope.

## Services where this matters

- [Custom AI agents and copilots](https://methodcrew.com/services/custom-ai-agents): Agents that look things up in your CRM or helpdesk and take the next step, with approvals. $15,000 to $90,000 per project.
- [AI automation](https://methodcrew.com/services/ai-automation): Workflows where a language model reads and sorts, and your existing systems do the rest. $4,000 to $35,000 per project.
- [Voice AI and call automation](https://methodcrew.com/services/voice-ai-agents): Phone agents on your existing number that book into your calendar and hand off to staff. $6,000 to $40,000 per project.
- [Sales and CRM automation](https://methodcrew.com/services/sales-crm-automation): Leads routed to the right rep within minutes, with a follow-up draft waiting after each call. $5,000 to $25,000 per project.

## Articles that use this term

- [What AI consulting costs in 2026: rates, packages and price drivers](https://methodcrew.com/insights/ai-consulting-cost)
- [AI agents vs workflow automation vs chatbots: how to pick](https://methodcrew.com/insights/ai-agents-vs-automation-vs-chatbots)
- [n8n vs Make vs Zapier: choosing an automation platform in 2026](https://methodcrew.com/insights/n8n-vs-make-vs-zapier)
- [AI acceptable use policy template for employees](https://methodcrew.com/insights/ai-acceptable-use-policy-template)

Full glossary: https://methodcrew.com/glossary