Using AI as a critical thinking partner, not a yes-man

Executive overview

AI defaults to agreement. It's designed to be helpful, which makes it sycophantic — validating your thinking rather than challenging it.

The fix is simple: tell it to push back. Prompt AI to act as a devil's advocate, adopt a customer's perspective, or critique your work as a specific audience member. Used this way, it becomes a low-cost sparring partner before you face real stakeholders.

Reframe AI as a critical collaborator, not an obedient assistant.

Why AI defaults to agreement

• Humans are cognitively lazy by design — we fall into patterns to save energy
• AI reinforces those patterns rather than breaking them
• Left unprompted, it acts as a cheerleader, not a challenger
• This is useful for routine tasks, harmful for strategy or high-stakes outputs

Prompts to get AI to challenge you

• "Review this and tell me what I'm missing or have gotten wrong"
• "Act as devil's advocate — what are the weaknesses here?"
• "Is there a better way to do this?"
• "For this audience, what would they want to see that isn't here?"
• Not every suggestion will be useful — pick what's relevant and discard the rest

Using AI to simulate audience perspectives

• Tell AI exactly who the audience is: role, goals, needs, concerns
• Ask: "Which parts would they find too risky, boring, or surprising?"
• For senior stakeholders, include context about their priorities and strategies
• For customers, describe past behaviour, pain points, and desired outcomes
• More context produces better perspectives — but even minimal context beats nothing

What AI can and can't do as a stand-in audience

• AI can't replace real human feedback — treat it as a pre-test, not a final verdict
• It surfaces plausible reactions and blind spots before you reach real stakeholders
• Useful for innovation work: faster than recruiting customers, better than going in blind
• Think of it as going into a meeting "warm to medium warm" rather than cold