ChatGPT prompt for running a lost-deal debrief

This is the prompt we run within a week of losing a deal we cared about, to pull a lesson from it without rewriting history. Its core rule: it works from your notes alone, marks the gaps [UNKNOWN], and labels speculation as speculation instead of dressing it up as fact.

The prompt to copy

We lost a deal. Context: {who, what we proposed, price, timeline of the deal, their stated reason}.
Write a debrief:
1. Timeline reconstruction — from my notes only; mark gaps [UNKNOWN].
2. Likely real reason(s) vs stated reason — label anything speculative as speculation.
3. What was in our control vs not.
4. One process change worth making — only if the evidence actually supports one; "no change warranted" is a valid answer.
5. A short, gracious close-out email that keeps the door open, under 90 words.

Replace the field in curly braces with your context: who, what you proposed, the price, the timeline, their stated reason. Paste your notes exactly as they are.

When to use it

In the days after a loss, while the details are fresh but the sting has faded. A debrief run too hot turns into a trial or a round of self-blame. This prompt imposes a discipline: what your notes prove on one side, what you are guessing on the other, and a single question at the end, was a process change really worth making.

What you get back

1. A timeline reconstructed from your notes alone, gaps included.
2. The likely real reason, separated from the stated one and labeled as speculation.
3. The split between what was in your control and what was not.
4. One process change, only if the evidence supports it.
5. A gracious close-out email, under 90 words, that keeps the door open.

Real example: loss notes → debrief

Input: the Brightside agency, an 8-person shop, has just lost a range refresh to a competitor of Fernhill Coffee. Lena’s notes, as-is:

proposed 5-label refresh, price sent on the 3rd, follow-up on the 10th got no reply
call on the 14th: "we're going with someone in-house"
not sure price was the blocker, nobody said so
timeline? they may have wanted it faster than our 6 weeks

Output, the part that matters:

2. Likely real reason vs stated reason Stated reason: hiring someone in-house. Likely real reason: [speculation] the 6-week timeline may have weighed, given the prospect hinted at an earlier launch, but no note confirms it. Price appears nowhere as a blocker: filing it as the cause would be a guess, not a fact.

3. What was in our control The follow-up on the 10th went unanswered for four days before their call: a checkpoint we can’t place [UNKNOWN], we don’t know whether they had already decided by then.

Notice what the AI did not do: it did not rule that price killed the deal when nobody said so, and it marked [UNKNOWN] the point where the decision actually turned. That is the prompt working, not failing. The human checkpoint, judging whether a change of method is truly warranted, stays in your hands.

The full worked example (complete notes, full debrief, close-out email) ships inside the kit, with the other prompts in the sales phase.