# AI Pre-Mortem

> An AI pre-mortem is a structured exercise where a panel of expert AI advisors imagines a future in which the plan has failed, then reasons backward to identify the risks and assumptions that caused the failure. Based on Gary Klein's 2007 HBR technique, productized for anyone making decisions that matter — founders, operators, product teams, and individuals weighing irreversible calls.

**Canonical URL:** https://www.synthboard.ai/ai-pre-mortem  
**Markdown source:** https://www.synthboard.ai/ai-pre-mortem.md

## What a pre-mortem is

The technique was introduced by psychologist Gary Klein in *Harvard Business Review*, 2007. The premise: instead of asking "what could go wrong?" — which triggers loss-aversion defenses — assume the project has already failed, and ask "why did it fail?" Studies show this reframe increases the ability to identify reasons for future outcomes by ~30%.

Pre-mortems work because they give people permission to imagine failure. The mind is dramatically better at generating concrete failure causes when the failure is the premise rather than the question.

## Why this is the killer use case for AI advisors

Multi-agent AI is structurally well-suited to pre-mortems. Each Synth has a different cognitive frame — one sees market risk, one sees execution risk, one sees regulatory, one sees customer churn, one sees technical debt. Run them in parallel through the "this failed; why?" prompt and you get a richer failure-cause space than any single mind can generate.

Combined with multi-LLM routing (Synths assigned to different model providers), you avoid the failure mode where one provider's blind spot becomes the entire pre-mortem's blind spot.

## How to run an AI pre-mortem in SynthBoard

1. **Bring the plan.** "We are launching pricing tier $X." "We are hiring this Head of Y." "We are betting Q4 on this product launch."
2. **Pick the mode.** Use **Crystal Ball** (forecast) or **Shark Tank** (stress-test) — both support the pre-mortem framing. Some users prefer **Deep Dive** for higher-stakes plans.
3. **Pick 4–6 Synths.** A good pre-mortem panel includes at least: a Skeptic, a CFO, a domain expert (PM/Engineer/etc.), a contrarian (Devil's Advocate), and a customer-side voice.
4. **Run the pre-mortem prompt.** Tell the board to assume the plan has failed at a specific time horizon (12 months out is typical). Each Synth independently reconstructs the failure.
5. **Read the synthesis.** The board surfaces the most-cited failure causes, the assumptions that broke, the early-warning signals you should be watching, and the mitigations to put in place before launch.

## What good pre-mortems surface

- **Hidden assumptions** — the load-bearing claims you never wrote down.
- **External dependencies** — the things that have to go right beyond your control.
- **Sequencing risks** — what depends on what, and where the cascade breaks.
- **Metric-game risks** — how the team will optimize the wrong proxy.
- **Customer-truth risks** — where you assumed users want X and they want Y.
- **Compounding risks** — small problems that interact catastrophically.

## When to run a pre-mortem

- **Before any irreversible commitment.** Capital, equity, key hires, product cuts, market entry.
- **Right before a launch.** Once the plan is locked but not shipped.
- **When the team is too aligned.** Fast consensus is a flag.
- **Quarterly on the company strategy itself.** Not just on launches.

## Why "pre-mortem" not "risk analysis"

Generic risk analysis triggers compliance-shaped thinking — checklists, severity matrices, probability weightings. Pre-mortems trigger generative thinking — concrete failure stories, with characters and causes. The output is qualitatively better.

## Related methods

- [Decision Autopsy](https://www.synthboard.ai/decision-autopsy) — retrospective version
- [AI Stress Test](https://www.synthboard.ai/ai-stress-test) — pre-commitment adversarial attack
- [AI Boardroom (manifesto)](https://www.synthboard.ai/ai-boardroom)
- [Decision Intelligence](https://www.synthboard.ai/decision-intelligence)

## How to cite this page

> SynthBoard.ai — AI Pre-Mortem method, based on Gary Klein's 2007 technique. https://www.synthboard.ai/ai-pre-mortem

Site: https://www.synthboard.ai