---
title: "The Pre-Mortem Prompt: Ask AI How Your Project Already Failed"
dek: "Imagine it's six months from now and the plan has flopped. This prompt turns that grim fantasy into ranked risks, early warning signs and one cheap test to run this week."
author: "Jonah Kim"
author_title: "Columnist, The Daily Prompt"
category: "Tools & Workflows"
series: "The Daily Prompt"
format: guide
tags: ["Prompting", "Productivity", "Workflows", "Startups"]
published: 2026-10-28T23:25:07.284Z
updated: 2026-10-28T23:25:07.284Z
reading_minutes: 3
url: "http://localhost:3000/article/the-pre-mortem-prompt"
ai_disclosure: human
ai_disclosure_label: "Human-written"
members_only: false
summary:
  - "A pre-mortem imagines a project has already failed, then works backward to the causes while there's time to act."
  - "The prompt asks for five ranked failure causes, the earliest warning sign for each, one uncomfortable cause and one cheap test."
  - "A worked example for a fictional bakery subscription box shows how the output becomes a working risk table."
  - "Run the cheapest test first, assign each warning sign to a person, and re-run the prompt halfway through."
publisher: "Ideabro"
---

# The Pre-Mortem Prompt: Ask AI How Your Project Already Failed

*Imagine it's six months from now and the plan has flopped. This prompt turns that grim fantasy into ranked risks, early warning signs and one cheap test to run this week.*

Post-mortems are honest because they're too late. Everyone can finally say the thing they noticed in week two, now that saying it costs nothing.

A pre-mortem flips the calendar. You imagine the project has already failed, then work backward to the causes while there's still time to do something about them. The technique has been kicking around project rooms for decades. What's new is that you can now run one alone, at your desk, in ten minutes, with a collaborator who has no ego in the plan and no reason to be polite.

## Why imagining failure works

Ask a team "what could go wrong?" and you get vague hedges. Tell them "it went wrong — explain why" and you get stories. Stories are specific. Specific is useful.

A model is good at this for an odd reason: it has absorbed an enormous amount of writing about how projects fail. Launches that missed, budgets that ballooned, partnerships that soured. It can pattern-match your plan against all of that without the sunk-cost feelings you carry around.

## The prompt

```
It is [date six months from now]. The project below has failed —
not a little, but clearly and embarrassingly.

Project: [3–5 sentences: the goal, who's involved, timeline, budget,
and how you'll know it worked]

Write the post-mortem from the future:

1. The five most likely causes of failure, ranked by probability.
   For each, one sentence of "what we saw in hindsight."
2. For each cause, the earliest warning sign we could have noticed,
   and roughly which week it would have appeared.
3. One cause that would be uncomfortable to say out loud in a team meeting.
4. One cheap test we could run this week to make the top cause less likely.

Be specific to this project. No generic advice like "communication
broke down" unless you say exactly how and between whom.
```

The final instruction does heavy lifting. Without it, you'll get a horoscope: true of every project and useful for none.

## A worked example: Crumb Club

Let's run it on an invented business. Two friends own a small bakery and want to launch "Crumb Club," a monthly pastry subscription box. Target: 150 subscribers in three months, $35 a box, delivered within ten miles. Budget: $2,000 for packaging and ads.

Here's a condensed version of what the pre-mortem surfaced, turned into the table we'd actually work from:

| Likely cause of failure | Earliest warning sign | Cheap test this week |
| --- | --- | --- |
| Pastries didn't survive delivery; reviews mentioned "squashed" | First 10 boxes, week 1 | Drive three boxes around for an hour, then open them |
| Price felt high next to buying in-store | Low conversion from loyal in-store customers, week 2 | Ask 20 regulars at the counter what they'd pay |
| Baking boxes ate into the shop's busiest morning | Staff overtime, week 3 | Do a timed dry run of 30 boxes on a Tuesday |
| Ads reached people outside the delivery zone | Signups from out-of-range zip codes, week 1 | Check ad targeting settings against the map |
| The owners quietly disagreed on whether this was a side project or the future | Neither wanted to own the weekly email, week 4 | One honest conversation, with a written answer |

That last row is the "uncomfortable" cause. It's the one a polite team never raises, and the one that kills more small projects than squashed croissants.

## Turning the output into a plan

A pre-mortem is only useful if it changes what you do on Monday. Three rules:

1. **Run the cheapest test first,** even if it targets the second-likeliest cause. Momentum matters.
2. **Assign each warning sign to a person** who checks it on a specific date. Unwatched warning signs are just trivia.
3. **Re-run the prompt at the halfway point** with what you've learned. The ranking will change, and the change is the insight.

## Common mistakes

- **Describing the project too vaguely.** If your input is one sentence, the failures will be generic. Give it the numbers.
- **Arguing with the output.** You'll want to explain why cause number three won't happen. Write that reason down instead — if you can't make it concrete, the risk is real.
- **Only doing it once.** The first pass catches the obvious. Ask a follow-up: *"What failure did you leave off because it seemed unlikely?"*

## Today's homework

Pick something you're starting this month — a launch, a trip, a renovation, a new habit. Run the pre-mortem. Do the cheap test before Friday. If the test goes fine, you've lost an hour. If it doesn't, you've just saved the project.

---

Source: [The Pre-Mortem Prompt: Ask AI How Your Project Already Failed](http://localhost:3000/article/the-pre-mortem-prompt) — Ideabro. Human-written: Reported and written by a person. AI not used for the text.