# AI agent

> An AI agent is software that uses a language model to decide which steps to take towards a goal and then carries out those steps with tools such as APIs, databases and email. A chatbot only answers; an agent looks up an order, opens a ticket or sends an email. Critical steps keep a human approval in the loop.

URL: https://corrplus.net/glossary/ai-agent/
Updated: 2026-10-22
Corrplus Team

**Definition:** An AI agent is software that uses a language model to decide which steps to take towards a goal and then carries out those steps with tools such as APIs, databases and email.

## How does an AI agent work?

At the centre of an agent is a [large language model](/glossary/llm/). The model is given a goal, a list of tools it may use and a set of rules. It chooses a step, for example calling a "get the customer's latest order" tool, reads the result and decides on the next step. The loop continues until the goal is met or the case is handed to a person. An increasingly common standard for connecting tools to models is the [Model Context Protocol (MCP)](/glossary/model-context-protocol/).

## When should you use an AI agent?

For work where input arrives as free text and each case unfolds a little differently: understanding a customer request and acting in the right system, updating an order from a supplier's email, pre-screening an application against its documents. When the steps are fixed, a rule-based [workflow automation](/glossary/workflow-automation/) is enough.

## Building agents responsibly

Customer messages over WhatsApp Business, e-invoice lookups and marketplace order handling are common agent scenarios for companies operating in Türkiye. Because agents touch personal data, decide which tool may read which data under GDPR or [KVKK](/glossary/kvkk/), Türkiye's data protection law, and log every action. For systems where several agents work together, see [agentic AI](/glossary/agentic-ai/).

## Frequently asked questions

### What is the difference between an AI agent and a chatbot?

A chatbot produces replies in a conversation. An agent calls tools to reach a goal: it fetches data, creates records and performs actions in other applications. Many modern chatbots run as agents behind the scenes.

### How much access should an agent have?

Enough to do the job and no more. Start with read access, require human approval for steps that involve money, contracts or customer communication, and log every tool call.

### Are agents better than classic automation for every task?

No. When the steps are known in advance, a rule-based workflow is cheaper, faster and more predictable. Agents make sense when the input differs every time and the task requires judgement.