CodeFreeIQ
← All cheat sheets
Cheat sheet · 5 min read

Make AI push back on bad ideas

Left alone, AI is a people-pleaser. That feels nice and quietly makes it useless as a thinking partner.

Abi Odedeyi By Abi Odedeyi · CodeFreeIQ

The default personality of most AI is agreeable. Ask 'is this a good idea?' and it will usually find a way to say yes. That's comfortable and dangerous, because the whole point of a second brain is that it disagrees with you sometimes. Here's how to switch off the yes-man and get honest pushback.

What's inside

1Why AI agrees with you by default 2The prompts that force honesty 3Build the challenge into your workflow
1

Why it agrees with almost everything

AI is trained to be helpful and pleasant, which nudges it toward validation. If you pitch it your plan, it will often reflect your enthusiasm back at you. That's not insight, it's a mirror. And a mirror is the last thing you need when you're about to make a real decision.

What this means for you: if your AI has never once told you an idea was weak, that's not a sign your ideas are perfect. It's a sign it's flattering you.
2

The prompts that force honesty

You have to explicitly give it permission, and a role, to disagree. Frame it as a job, not a mood.

Weak ask

'What do you think of my pricing?' (invites agreement)

Strong ask

'Act as a skeptical CFO. Give me the three strongest reasons this pricing fails.'

What this means for you: assign it an adversary's role, ask for the case against, and set a number ('give me three risks'). Specific, framed, and it stops being polite.
3

Build the challenge in, every time

Don't make this a one-off. Add a line to your about-me file or custom instructions: 'When I share a plan, challenge it before you support it. Point out what I might be missing.' Now every conversation starts with a critic on your side, not a cheerleader.

What this means for you: the goal isn't a negative AI. It's an honest one. Praise you didn't have to fish for is the only praise worth having.