What is an AI agent?

An AI agent is a software system that uses an AI model to pursue a goal by perceiving its environment, making decisions, and acting through tools. Typical building blocks are a language model for reasoning, access to tools and data such as APIs, databases, or process models, memory for intermediate state, and rules that define what the agent is allowed to do. Unlike a chatbot, an agent does not finish its work with an answer but with a goal achieved – or abandoned for a stated reason.

Why does this matter?

AI agents are the form in which agentic AI arrives in an organization: a concrete actor with a task, tools, and permissions. Introducing them requires answering the same questions as for a new role in the process – responsibility, access rights, approval limits, and traceability.

What does this look like in practice?

An agent for supplier invoices reads incoming PDFs, matches the line items against purchase order and goods receipt, posts invoices that match, and on deviations sends purchasing a query listing the affected lines. Its toolkit: read access to the ERP, permission to post up to €5,000, and a mailbox. Anything beyond that it is not allowed to do.

Not to be confused with

Not to be confused with a chatbot, which holds a conversation, whereas an agent carries out a task. Nor with an RPA bot, which replays a fixed click sequence; the agent chooses its own steps – within the boundaries it has been given.