How should a person approve an action prepared by an agent?
Review note: The answer refers to the original author's CV or target job offer — adapt it to your own experience.
Answer out loud or in your head first, then compare.
Answer
First the agent prepares a specific proposal: what to change, in which object and with what parameters. The user approves that version, not a general consent to anything the model comes up with later. Before execution, the backend checks the person, their permissions and that the proposal is still current.
If the amount or recipient changed after approval, a new approval may be needed. A double click must not perform the operation twice. That's the practical role of human-in-the-loop, i.e. bringing a person in at an important point in the process. A button in Teams is the interface for this approval, but on its own it doesn't replace backend checks and recording who approved what.
Why it matters: The concrete meaning of the HITL requirement in the offer.
Possible follow-ups: What if the user clicks an old card the next day?
How did it go? Be honest — “Not yet” puts the question on your review list.
Start with
- H-P01 · How does an AI agent differ from a plain model call? — Agents and workflows
- H-P02 · What is a workflow, i.e. a fixed sequence of work? — Agents and workflows
Practise in code
- Approving a specific proposal — runs tests in your browser
- An assistant with a new requirement — guided self-review
Further reading
- Microsoft Graph: Overview of permissions(opens in a new tab)Docs; accessed 22.10.2026. Delegated and application permissions, and the limits of access.