AI tools are trained to be agreeable. Left alone they will tell you your plan is strong, your numbers look healthy and your idea has real potential. That is worth nothing. These are the five rules I use every day to get straight answers instead.
Never ask "is this good?"
Ask "what's wrong with this?" instead. Same document, same numbers, completely different answer. The first question invites praise. The second gives it permission to be honest.Make it argue against you.
Before any decision, ask for the strongest case for not doing it. If the case against is weak, proceed. If it makes you wince, you have just saved yourself money. I do this with supplier deals, hires and pricing.Ask what a skeptic would say.
"Read this as my most cynical competitor. What would they pick apart?" This catches what your own optimism hides. Works on emails, proposals and anything you are about to publish.Tell it to stop agreeing with you.
Say it plainly: "Don't flatter me. If this is average, say average." You have to give it permission. It defaults to polite.Never let it have the last word on numbers.
AI is most confident exactly when it is wrong, and it is most wrong with figures. Anything with a pound sign gets checked against the source before it goes near a decision. Treat it like a bright new starter: fast, useful, never left unsupervised.
The pattern behind all five: by default, AI is a mirror. These rules turn it into an advisor. An advisor that always agrees with you isn't one.
More of this every week. You're already in the right place.