Most people prompt the AI, accept the first answer, and move on. The people getting real value are doing something different: they're forcing the AI to pressure-test its own thinking, restate the problem in its own words, and surface the assumptions it's quietly making. Tonight, we walk through five specific prompting moves — restate, critique, constrain, compare, and challenge — that change the quality of the output more than any clever phrasing ever will.
You'll run all five on a single real decision — start to finish, in one chat. Bring your own if you have one. Tonight's running example: a classic build-vs-buy call.
Before trusting any answer, make the AI play the problem back in its own words — and name the assumptions it's making. Half of bad outputs come from a misread question, not a bad model.
The highest-leverage move. Once it gives a recommendation, make it find the holes — as if a skeptical expert were reviewing it. The self-critique is often more useful than the original answer.
Vague questions get vague answers. Add the constraints that actually bind you — budget, timeline, headcount — and make the AI trade against them, not wish them away.
A single recommendation has nothing to push against. Make the AI lay its pick next to the real alternatives so you can see the tradeoffs, not just the winner.
The last move, and the one people skip. Make the AI build the strongest case AGAINST its own recommendation — and tell you what would change its mind. If it survives its own counterargument, your confidence is earned, not borrowed.
The slides, the participant guide, and the live Q&A one-pager — everything you need to bring the five pushback moves into your own work. Drop your email to unlock the downloads.
Slides, the full prompt pack, and the one-pager Q&A recap — every question asked in the room, answered. Plus the occasional note from Momentum on what we're building next.
The full presentation — every slide, every move, every prompt we walked through in the room.
Download slidesThe full pushback prompt pack — copy-paste-ready, with the five moves, what good looks like, and when to push back on the AI's answer.
Open the guideEvery question asked in the room — answered. Plus the five-move framework and the confidence traps to watch for, on a single page.
Download one-pagerBuilt live with the room. The five pushback moves packaged as a Claude skill — drop it in, call "apply the adversarial review" from any chat, and it runs the whole framework on any decision.
Download skillGreat workshop. Eye-opening.— Prompts That Push Back attendee, June 2026
Based on 10 post-workshop survey responses.
Two minutes of feedback shapes the next Power Hour more than anything else. Tell us what worked, what didn't, and what you'd want next time.
Share feedbackJoin the Momentum Product Co. Slack to share what you tested, compare pushback checklists, and see how other folks are getting their AI to argue with itself.