Why this matters
Ask people "what could go wrong?" and they list a few safe, polite risks. Tell them "it's six months from now and this failed, so write down why," and they come up with far more specific reasons. Product teams call this a pre-mortem: a post-mortem you run before the project starts, when the answers can still change the plan.
AI is a useful pre-mortem partner because it has no ego in your plan. It can write the failure story your team may be too polite to say out loud, and that story is often more useful than a generic list of risks.
When to run one
- Before a launch, a product build, or a big feature.
- Before a job change, a move, or a big purchase.
- Before you commit a quarter of your time to anything.
- Any time everyone in the room agrees a little too fast.
The pre-mortem prompt
Describe your plan and its constraints. The prompt asks for distinct failure stories, early signals, and changes you can make before committing.
Run a pre-mortem on my plan. It's [date six months from now], and the plan has clearly failed. Your job is to explain why, as specifically as possible, so I can prevent it.
THE PLAN
- What I'm doing: [describe the project, launch, or decision]
- The goal: [what success looks like, with a number or date]
- Who's involved: [me alone, a team, partners, customers]
- Resources: [time, budget, people]
- What I'm most confident about: [the part you think is solid]
STEP 1: THE FAILURE STORIES
Write 5 different stories of how this failed, each about 80 words, told as if looking back. Make them specific: what happened, when it started going wrong, and the moment it became clear. Cover different kinds of failure:
- A market or customer failure (nobody wanted it, or the wrong people did)
- An execution failure (it took too long, cost too much, or broke)
- A people failure (someone left, disagreed, or lost interest, including me)
- An outside failure (a competitor, a platform change, the economy, a rule change)
- A failure in the part I'm most confident about
STEP 2: THE EARLY SIGNALS
For each story, name the earliest warning sign I could have spotted, and when I'd likely see it.
STEP 3: RANK THE RISKS
Rank the 5 by how likely they are and how bad they'd be. Show a table: story / likelihood (low, medium, high) / damage (low, medium, high) / early signal.
STEP 4: CHANGE THE PLAN
For the top 3 risks, give me:
- One change to the plan that makes it less likely
- One stop signal: a specific, measurable signal and a date to check it, that tells me to stop or change course
- What I'd do if that signal shows up
RULES
- Don't reassure me. Every story should be realistic, not dramatic.
- If my goal or timeline already looks unrealistic, say so before the stories.
- Use what I told you. Mark any assumption you made with [ASSUMED].
Run it with a team
Have everyone write their own failure story first, without AI, in five quiet minutes. Then run the prompt and compare. The stories AI finds that nobody on the team wrote are the blind spots, and the ones everyone wrote are the risks you've been avoiding saying out loud.
Pro move: calendar the stop signals
Put the stop signals on your calendar on the day you write them. A pre-mortem only pays off if someone actually checks the signal in month two.
Do this next
Pick the biggest thing on your plate this quarter and run Step 1 only. Read the five stories and notice which one made you uncomfortable. Start there.