Curriculum

Cluster 6 ¡ Lesson 3 1 min read

The Decision Log

Why documenting what you tried, what worked, and what did not is the most valuable output.

The Decision Log is arguably the most important artifact you will produce in the AI Process Lab. It is the mechanism by which individual experimentation is transformed into institutional knowledge. Without it, the lessons learned from what worked—and more importantly, what didn't—will simply evaporate in the endless scroll of Slack channels and forgotten emails.

By rigorously documenting your hypotheses, outcomes, and the reasoning behind your decisions, you create a 'rejection library.' Every failed prompt, every hallucinated output, and every necessary human intervention becomes a labeled counterfactual. This log is not a testament to failure, but a high-fidelity training signal that teaches your team the boundaries of your tools and refines your collective approach to AI integration.

Assignment

Create a Decision Log for a recent task where you used an AI tool. Document the following:

  1. What you tried to do.
  2. Your initial hypothesis (what you expected the AI to produce).
  3. What actually happened (the output).
  4. What you decided to do with that output (edit, reject, use as-is).
  5. Why you made that decision. Share this log with a colleague and discuss how it might help them in a similar situation.

Learning Objectives

  • Understand why the Decision Log is the most critical artifact of the AI Process Lab.
  • Learn to structure a Decision Log to capture hypotheses, outcomes, and reasoning.
  • Recognize how failures and rejections serve as valuable training signals for future work.

The Decision Log

A structured record of what was tried, the underlying hypothesis, the actual outcome, the final decision, and the reasoning behind it. It prevents institutional knowledge from being lost in ephemeral channels like Slack.

The Rejection Library

Inspired by an INSEAD study, this concept treats every rejection, edit, or reroute as a labeled counterfactual. These 'failures' teach boundaries and refine future approaches.

The binding constraint on AI adoption is not technical skill, but the ability to see your own work clearly enough to redesign it.