Vic Sol · Decision DeskOne-page sheet · 15 min to use
Pre-mortem: the decision already failed, why?
For the team about to approve a hard-to-reverse AI decision that wants to hear the risks before signing, not twelve months later.
When to use it
Tomorrow you approve an AI purchase, contract or rollout that would take months to undo.
Everyone nods in meetings and nobody objects, and that worries you more than an argument would.
There was already a project "nobody saw coming" and you do not want to hear that sentence again.
The sheet
Instruction for the group: "It is twelve months from today. The decision we are about to approve failed. Not partially: we cancelled it. Write down why."
Reason it failed
Early signal (what we would see at 60–90 days)
What we would change today
Owner
[One concrete cause, in one sentence. Illustrative: "The sales team never adopted the tool."]
[An observable data point, not a feeling. Illustrative: "Fewer than 20% of reps open it in week 8."]
[One action before approval. Illustrative: "A named rep joins the pilot from week 1."]
[Name]
Decision under review: ____________________ · Pre-mortem date: ____________ · Next signal check: ____________
How to fill it in 15 minutes
Minutes 0–2. Read the instruction aloud. Nobody speaks yet. "It already failed" is not the facilitator's opinion; it is the rule of the exercise.
Minutes 2–6. Silent writing. Each person writes at least three reasons on their own sheet, no comments. Silence protects the uncomfortable reasons.
Minutes 6–10. Round. Each person reads one reason, in turn, with no reply. Repeat until no new reasons remain. Nobody defends the decision during the round; that moment comes later.
Minutes 10–14. Group the repeated reasons and pick the four or five that worry you most. For each one, fill in the early signal and what you would change today.
Minute 15. A name in every row. A row without an owner is a risk nobody is watching.
The most common mistake: using the exercise to confirm the decision is sound. If few reasons come out, or they are all mild, the group is protecting whoever proposed the decision, not the company.
Rules for running it
No ranks during the round: the newest person reads before the most senior one.
Reasons are written before anyone speaks. Speaking first makes everyone write the same thing.
Do not debate whether a reason is "likely"; debate which early signal would give it away.
Whoever proposed the decision takes part, writes and listens. They do not facilitate.
The line that is almost always missing
The early-signal column. A pre-mortem that ends in a list of risks gets filed. A pre-mortem that ends in an observable signal at 60 or 90 days, with an owner, becomes an item on the next review agenda. That column is the difference between an exercise and a control.
Vic Sol · Decision Desk · vicsol.org/resources/ · Free to use with attribution. This sheet structures a conversation; it is not legal, financial or audit advice.
Open the .docx in Word (or the .xlsx in Excel) or print it before the meeting.
Fill the fields with the team; leave the blanks visible, they are the discussion.
Close with one owner, one date and one next action. Nothing else.
Prompt
Generate this sheet for your case
Paste this prompt into ChatGPT, Claude or Gemini together with your proposal, process or policy. It reproduces the sheet with your data and flags what is missing.
Act as a chief operating officer with 15 years of experience who has watched technology rollouts fail for reasons nobody said out loud in the approval meeting.
Context I am giving you:
- Company and area: [e.g. distributor, 400 people, sales area]
- Decision about to be approved: [one sentence: what, for whom, how long it commits us]
- What makes it hard to reverse: [contract, integration, data, people]
- Who proposes it and who approves it: [roles]
- Proposal or memo, if any: [paste text or write "none"]
What I need: assume it is twelve months from now and the decision failed completely. Produce a pre-mortem with exactly this structure:
1. Eight reasons for the failure, each in one concrete sentence. Cover different angles: adoption, data, vendor, process, people, cost, customer, compliance.
2. For each reason, an early signal observable at 60–90 days (a data point, not a feeling).
3. For each reason, what we would change today, before approving.
4. The three reasons that, given the context, should worry us most, and why.
5. Three uncomfortable questions the group will probably avoid and that are worth asking in the session.
Output format: a table with the columns Reason · Early signal · What we would change today · Suggested owner (role, not name), followed by points 4 and 5 as bullets. Maximum 450 words. No introduction.
Rules:
- Do not invent figures or other companies' results. If a signal needs a threshold, write [THRESHOLD PENDING] and tell me how to set it.
- Mark every assumption as an assumption.
- No generic reasons ("poor communication"); each reason must name what failed and where in the process.
- If critical information is missing, ask me at most 3 questions before answering.
- Answer in plain, direct business English, with no sales adjectives.
What to check in the answer:
Is each early signal something you can measure or see on a date, or did it slide back to "the team is not engaged"?
Is one of the eight reasons the one your team does not dare to say? That one goes first in the session.
Do the "what we would change today" items fit before approval, or are they tasks for later? If later, they are not useful.
Vic Sol