The fastest way to a bad course is to turn an SME transcript straight into slides. These three prompts pull the decisions out of the content first — you end up editing judgment calls instead of formatting bullets. Run them in order; paste your material where marked.
1 — Find the moments that matter
Here are my notes from a subject-matter interview:
[PASTE NOTES]
List the 3–5 moments where someone on the job has to make a judgment
call and could plausibly get it wrong. For each, give: the situation,
the right move, and the tempting wrong move people actually make.
Ignore anything that's just reference information.2 — Draft a decision-first scenario
Take this moment:
[PASTE ONE MOMENT FROM STEP 1]
Write a short scenario where the learner faces the decision before any
teaching. Give one realistic setup, three options (one right, two that
reflect real misconceptions), and one-line feedback for each that
explains the reasoning, not just "correct" or "incorrect."3 — Make it measurable
For that scenario, propose the xAPI statements it should emit (launched,
answered with the choice, passed/failed) and the single on-the-job
behavior it maps to, so it can be tracked to Kirkpatrick Level 3
instead of just completion.Why decision-first beats content-first
A transcript is organized by what the SME knows; a scenario has to be organized by what the learner decides. Prompt 1 does that inversion — it discards reference material (which belongs in a job aid, not a course) and keeps only the moments with a tempting wrong move. If prompt 1 returns fewer than three genuine judgment calls, that’s a signal the content wants to be a job aid or a process fix, not training.