Glossary · AI agent
What is an AI agent?
Definition: An AI agent is a system that uses a language model to pursue a goal, taking steps with the data and tools it has been given permission to use.
An agent combines three elements: a language model that understands and writes text, a set of tools it can use, such as searching documents, reading an ERP record or creating a task, and instructions that set out what it may and may not do. Unlike traditional automation, it can work with free text: a hastily written email, a scanned contract or a question phrased in any way.
In a business, agents work well on tasks with plenty of text, high volume and criteria that can be explained with examples: classifying emails and requests, extracting data from documents, answering internal questions with a reference to the source, or preparing drafts for a person to review. They do not replace the company's judgement; they apply it faster and more consistently.
A chatbot converses; an agent also acts on systems. That is why it needs limits: a narrow scope, minimum permissions, human approval for irreversible actions such as sending, paying or deleting, and a log of every decision. Its quality must be measured against real cases, because the model can produce answers that sound right but are wrong. Grounding answers in the company's own documents through RAG reduces that risk.
Example: an agent reads the job reports that technicians close, spots those that include billable work and prepares the invoicing request for the accounts team to approve.