People have a built-in flaw: once we form a plan, we keep finding reasons it’ll work — confirmation bias — until we’re sure. Only after sinking real money and time in do the holes we ignored show up.
Most people use AI to agree with them (“isn’t my idea great?”). The far more valuable move is the opposite: make it the toughest opponent, hunting for the deadliest flaws in your plan. Getting “attacked” by AI before you commit is far cheaper than getting attacked by reality after launch. The value isn’t whether it’s right — it’s that it forces you to see the angles you can’t.
When to use it
Before any hard-to-reverse call — switching jobs, starting a business, sinking money into a project, locking a plan — let AI play the harshest critic and pressure-test it.
How to do it
- Lay out your idea or plan, plus your goal and current resources
- Explicitly tell it to take the opposing side and list the deadliest objections and overlooked risks
- Sort its points: which are real problems, which are noise
- Patch each real weakness, or ask “so how should I change it?”
- Go harder: have it “critique again as different personas” (investor / competitor / family)
Weak vs strong
The left mostly earns you cheerleading; the right — demanding the opposing side, five failure modes, no comfort — makes it actually poke at cash flow, location and competition you’d rather not face.
Copy-paste prompt
My plan:【describe it, with your goal, budget and resources】. Play the most critical, realistic skeptic: list the 5 likeliest reasons it fails and the risks I’m missing, ranked deadliest first, each with a concrete fix. Don’t go easy and don’t hedge — I need real problems, not encouragement.
Worked examples
You get：Three angles, three different holes: the investor on returns, the competitor on how you’d get copied, family on your fallback. Three distinct blind spots at once.
You get：It shifts from tearing down to problem-solving, giving concrete options with trade-offs — so you can decide whether and how to proceed.
Level up
- Red then blue: have it critique as the opponent first, then switch to “supporter” to help you counter — hear both sides
- Focus one axis: aim it only at “where I’m most likely to lose money” or “the part most likely to break” for depth
- Rehearse your pitch: let it attack, then refine — you’ve pre-played the sharp questions real people will ask
Common mistakes
- Too little context — without goal, budget and constraints it stays generic and misses the real pain
- Treating its objections as verdicts — they’re angles worth weighing, not rulings; you judge what’s real and what matters
- Scrapping the whole plan after one round — the point is to strengthen, not scare you off; good plans are forged by critique, not killed by it
FAQ
Will it just nitpick for the sake of it?
Which AI is best for this?
Pro tip：Far more useful than asking AI to praise you. Make it a habit: before any big decision, run it past your devil’s advocate — ten minutes now can save months of detours.