Human-in-the-loop
Human-in-the-loop means a human has a fixed place in the process: the AI prepares, the human approves or intervenes before anything becomes final.
Not every step needs a human to see it. The art is choosing where it counts: a quote going to a customer, a credit note, a reply to a complaint. There the agent stages the result and a human decides with one look and one click. Routine steps without consequences the agent may finish itself.
This is also what the AI Act asks for decisions about people, and what customers expect. A well-designed agent makes the check easy: everything it did is visible, and it can be undone.
Other terms
Prompt
A prompt is the instruction you give an AI model: the question, the context and the rules. With an agent the prompt is largely fixed in advance, so it behaves the same every time.
Fine-tuning
Fine-tuning is additional training of an existing AI model on your own examples, so it masters a specific style or task better. For most companies it is unnecessary: a good prompt plus your own data (RAG) gets further.
Token
A token is the piece of text (roughly part of a word) an AI model counts in. It determines the price of every request and how much text a model can process at once.
Want to know what this means for your business?
Tell us where you lose time or margin. We show where AI pays back fastest, in plain language.