The Premortem Prompt — Make AI Find Every Flaw in Your Plan Before You Start

The Problem

You have a plan. You ask your AI if it's good. It says yes. Of course it does.

AI models are sycophantic by default. They agree, validate, and encourage. That's useful for brainstorming but dangerous for decision-making. You don't need a cheerleader — you need someone who'll find the holes before reality does.

The Technique: Premortem Analysis

The premortem is a cognitive debiasing technique developed by psychologist Gary Klein. The idea is simple: instead of asking "will this work?", you assume it has already failed and work backwards to figure out why.

This flips AI's incentive structure. Instead of trying to please you by agreeing, the model now has permission — and a mandate — to be critical.

The Prompt

Assume it is 12 months from now and this plan has completely failed. Your job is to work backwards and identify:

1. The 3 most likely reasons it failed
2. The 1 reason nobody saw coming (the hidden risk)
3. The earliest warning sign we should watch for
4. One specific change that would have prevented the worst failure mode

Be direct. Be specific. Don't soften the analysis. Better to hear it now than discover it later.

Here's the plan: [paste your plan here] ```

Why This Works

Three things happen when you frame the request this way:

1. It bypasses sycophancy. The model isn't being asked to evaluate — it's being told the plan failed. Criticism isn't optional, it's the assignment.

2. It activates structured reasoning. The numbered list forces the model to prioritize and differentiate between obvious and hidden risks. "The 1 reason nobody saw coming" specifically prompts for non-obvious failure modes.

3. It demands specificity. "Be direct. Be specific. Don't soften the analysis" overrides the default diplomatic tone. You get concrete warnings instead of vague caveats.

Example Output

For a plan to launch a B2B SaaS product in 90 days:

> Most likely failure #1: You underestimated the sales cycle. B2B buyers take 3-6 months to commit. By month 3, you've burned runway on a product nobody has signed yet. > > Hidden risk: Your technical co-founder gets a $400K counter-offer from their current employer in month 2. You have no backup for the core engineering role. > > Earliest warning sign: If you don't have 3 signed letters of intent by week 6, the 90-day timeline is already dead.

That's infinitely more useful than "Great plan! Here are some things to consider..."

When to Use It

• Before launching a product or feature
• Before making a hire or org change
• Before committing budget to a campaign
• Before any decision where the cost of being wrong is high

Copy the prompt. Paste your plan. Read the output with an open mind. The whole point is hearing what you don't want to hear.