---
title: "Lessons learned from an AI initiative (without blame)"
slug: lecciones-aprendidas-iniciativa-ia
lang: en
situation: comunicar
format: plantilla
minutes: 20
updated: 2026-10-19
---

# Lessons learned from an AI initiative (without blame)

> For the team closing, pausing or scaling an AI initiative that wants the company to learn something before everyone moves on to the next project.

## When to use it

- The pilot ended and the final report says "mixed results" without saying what changes.
- The initiative is about to scale and nobody has written down what went wrong along the way.
- Something was cancelled and the corridor conversation already has a culprit. This sheet is there to replace it.

## The sheet

Initiative: ____________________ · Status: close / pause / scale · Session date: ____________

| Field | What to write | Illustrative example |
|---|---|---|
| What we expected | The number or result the original proposal promised, copied word for word. | "Quotes from 3 days to 1 in 8 weeks." |
| What happened | The actual number, with a date. No adjectives. | "Week 8: 2.5 days. Week 12: 2 days." |
| Which signal we ignored | The data point that was in plain sight before it became obvious, and when it was first seen. | "Week 3: only 2 of 9 reps were using the draft." |
| What we keep | What worked and stays, even if the initiative changes. | "The quote template and the single reviewer." |
| What changes next time | One concrete rule for the next initiative, not an intention. | "No pilot starts without a user from the area from day 1." |
| Who owns the change | One named person and the date the rule takes effect. | "Laura (sales), from the next pilot, 1 March." |

## How to fill it in 20 minutes

1. Before the session, the lead fills in "what we expected" and "what happened" from the original documents. Those two rows are not debated: they are data.
2. In the session, each person writes in silence for four minutes which signal they think was ignored. Then they are read in a round, with no reply.
3. Pick the signal most people saw. That is the lesson; the rest go in an appendix.
4. "What we keep" comes before "what changes". A team that only lists mistakes stops proposing.
5. Write "what changes" as a rule that can be checked in the next initiative, and give it an owner and a date.
6. The most common mistake: turning the signal row into a row of people responsible. The question is which data point was in sight, not who should have seen it.

## How to run the 45-minute session

- Minutes 0–5: the facilitator reads the rules: we talk about decisions and data, not people; nobody defends themselves; whoever led the initiative listens first.
- Minutes 5–15: data. Read the rows "what we expected" and "what happened". Correct figures, not interpretations.
- Minutes 15–30: signals. Silent writing, round, choice of the main signal.
- Minutes 30–40: what we keep and what changes. One rule, not five.
- Minutes 40–45: owner and date. Without this, the sheet gets filed with the project.

## The line that is almost always missing

"Which signal we ignored." Closing reviews tell what happened and what would be done differently, and skip the middle step: which data point was available and not looked at. That row is what stops the same closure from repeating under another project name.

## If this becomes a big decision

If this review produces a decision that is hard to reverse and carries your name, the free tools at vicsol.org/decision-clarity/ help you structure it. They do not replace a formal review.

---

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.