10 prompt patterns that turn AI answers into checklists
Ask an AI a vague question and you get a vague essay. Ask it the right way and you get a short, numbered list of things to do, each one clear enough to tick off. Here are ten prompt patterns that reliably produce actionable answers, in ChatGPT, Claude, Gemini or Workbranch.
Every pattern below follows the same principle: tell the AI what shape the answer should take. Language models are good at following format instructions, and a numbered list of specific actions is much easier to act on than five paragraphs of advice.
1. The numbered-actions pattern
The simplest upgrade to almost any prompt.
"[Your question]. Answer as a numbered list of specific actions. Start each item with a verb and keep each to one or two sentences."
Starting each item with a verb ("Add…", "Remove…", "Ask…") forces concrete actions instead of observations.
2. The ranked-by-impact pattern
"Give me the [5] most effective ways to [goal], ranked by impact. For each, note the effort (low, medium or high)."
Ranking tells you where to start. The effort label helps you find quick wins at the top of the list.
3. The definition-of-done pattern
"For each step, add how I'll know it's done."
A step like "improve onboarding" can't be ticked off. "Improve onboarding. Done when new users reach the first project in under two minutes" can. Add this line to any plan prompt.
4. The constraints-first pattern
"I have [2 hours a week / no budget / one developer]. Within those constraints, what should I do to [goal]? Numbered steps only."
Without constraints, AI tends to suggest the ideal plan for a company you don't work at. Constraints make the list realistic.
5. The review-as-checklist pattern
For feedback on writing, code, designs or plans:
"Review this [document]. List each issue as a separate numbered item: what's wrong, where it is, and the exact change to make. Order by importance."
Separate items mean you can fix and tick them off one at a time, instead of re-reading a block of feedback.
6. The next-three-steps pattern
"I'm working on [project]. Here's where I am: [status]. What are the next three steps?"
When a full plan feels overwhelming, asking for only the next three steps keeps you moving and lets you check in again once they're done.
7. The split-the-big-task pattern
"Break [big task] into steps that each take under an hour. Group them into phases."
Time-boxed steps are easy to schedule and finish in one sitting. Phases give you natural checkpoints.
8. The decision-matrix pattern
"Compare [option A] and [option B] for [my situation]. Give a short table of trade-offs, then a recommendation and the three things I should do to act on it."
Comparisons often end with "it depends". Asking for a recommendation plus next actions turns analysis into a decision.
9. The open-questions pattern
"Before giving the plan, list anything you'd need to know to make it better, as numbered questions. Then give your best plan with your assumptions stated."
This surfaces hidden assumptions. Each open question is also a natural follow-up thread: answer it, and ask the AI to update just that part.
10. The follow-up-on-one-point pattern
Once you have a list, dig into a single item without derailing the rest:
"Only about point 4: explain how to do it step by step. Don't repeat the other points."
The last sentence matters. Without it, many models rewrite the whole list. Even better, ask follow-ups in a separate thread so the original list stays intact. More on that in why follow-up questions derail AI chats.
Combine them
The patterns stack. A strong all-purpose prompt for any plan looks like this:
"I want to [goal]. My constraints: [constraints]. Give me the 5 most effective steps, ranked by impact, each starting with a verb, with effort (low, medium or high) and how I'll know it's done. List any open questions first."
Then track the list
A great checklist is only useful if you work through it. In most chat apps the list scrolls away as the conversation continues. In Workbranch, each point in an answer can be ticked off, highlighted, or followed up in its own thread, and the chat can carry a due date, so the list you prompted for actually gets done.
Looking for more ways to keep AI work organized? Read inbox zero for AI chats.
Give your AI chats an inbox
Projects, folders, due dates, highlights and branching threads for every AI conversation.
Try Workbranch now $29.90/month · Cancel anytime · No install needed