AI agent
An AI agent is software that pursues a defined job by choosing tools, reading state and producing a scorable outcome - not a FAQ chatbot alone.
How it works
You define a job, permissions and a stop condition. The agent observes inputs, plans steps, calls tools (CRM, mail, docs, APIs) and returns a result you can score. Guardrails limit which tools and arguments are legal. High-impact actions often pause for human approval. Evaluation sets made from real cases catch regressions when prompts or models change.
Agents fail when the job is vague, tools are over-permissioned, or nobody owns monitoring. They differ from fixed automations because they can branch; they differ from chatbots because they act, not only answer.
At NEXOVE, custom agents start with one measurable job, an evaluation set and an operate plan. See the AI agents hub for scope, cost drivers and when an agent is the wrong answer.
Example
An agent drafts a CRM follow-up from a call transcript, attaches a structured summary, and waits for a human before sending external email.