The anatomy of a good delivery prompt
A five-part structure that produces checkable drafts instead of confident guesses, mapped onto delivery tasks.
A reliable delivery prompt is structured, not clever. Five parts do most of the work, and they map neatly onto delivery tasks.
1. Role. "You are a scrum master helping refine a backlog" primes the model toward clarity and risk, not marketing copy.
2. Context. Give the relevant facts and nothing sensitive it does not need. Grounded context is the biggest lever on a useful answer.
3. Task. State the verb: "Draft acceptance criteria for this story" or "Summarize this retro into three actions with owners."
4. Constraints. "Use only the notes I provided. Mark any assumption. If a story is too large to estimate, say so." Constraints are where you stop the model from inventing.
5. Output format. "Return a table: Story | Acceptance criteria | Open question." A defined format makes the draft easy to check and drop into your tool.
The habit that matters most is asking the model to mark its assumptions and flag gaps, so the draft arrives ready for the team to question rather than pretending to be finished.
Integrity angle. A well-formed prompt produces a draft the team can inspect. A vague prompt produces confident text that hides what the model guessed. Structure keeps the human in control.
Example prompt: "You are a delivery lead. Using only the notes below, draft two user stories with acceptance criteria. Return a table and add a column for assumptions you made. Notes: [paste]."
Take a task you would normally hand AI in one line. Rewrite it using all five parts, and add the instruction to mark assumptions. Compare the two outputs.
Enjoying the free lessons? Get an email when we publish new courses and updates — no spam, unsubscribe anytime.
Discussion (0)
Ask a question or share what worked for you. Comments are reviewed before they appear.
No comments yet. Be the first to start the discussion!