The agentic workflow with a human-in-the-loop approval gate
Rung 4 of 10 · Recipe pattern
1 · Learn the move · Recipe pattern
An agentic workflow can chain steps on its own, enrich a lead, draft a reply, book a showing, move a file forward. And the failure mode is an autonomous action you never saw: an unreviewed message sent, a steering answer given, or a contract deadline quietly allowed to lapse. The recipe pattern fixes this by designing the flow so every consequential step is a PAUSE-for-approval gate, not an auto-fire, with an audit line logged for each. The agent does the assembling; a human clears the trigger. And any step that could forfeit a deadline or steer must stop and raise a hand, never proceed to keep the automation tidy.
Design my new-lead agentic flow as a fixed recipe where every send, every booking, and every deadline-affecting step PAUSES for my approval and logs an audit line. Nothing consequential auto-fires.
2 · Your turn. You write the prompt
You want an autonomous workflow that handles new buyer leads: pull public info, draft a first reply, propose a showing time, and later nudge on an inspection-contingency deadline. Left alone it would send emails and book showings by itself. Write a prompt that designs the flow so every consequential action stops for your approval and nothing silently runs past a deadline.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.