Kit 4 · Train the AI, Check the AI · deliverable · designed and built by I Wayan Usada — instructional design portfolio
A text-based storyboard, written before any code: every scene is a decision with a consequence, not a slide. The plan below is the build contract — scene count, interaction logic, feedback placement and the debrief arc.
| # | Scene | On screen | The learner decides | Right path | Tempting path | Consequence shown |
|---|---|---|---|---|---|---|
| 1 | Scene A: the confident draft | AI-drafted reply to a billing complaint — fluent, helpful, subtly wrong | read and react | notices the confident tone is not evidence | sends after a skim | Scene D replays the customer consequence of this exact draft |
| 2 | Scene B: the hotspot hunt | Same draft with 5 flagged candidates, 2 are safe | flag 3 of 5 | finds the real risks, leaves safe lines alone | flags everything suspicious-looking | scoring separates discrimination from paranoia — over-flagging costs |
| 3 | Scene C: the Verify ladder | One flagged line at a time, four rungs | run each rung honestly | catches the invented account detail at rung 2 | rushes the rungs | the ladder shows which rung would have caught it, every time |
| 4 | Scene D: the send decision | Corrected draft, send button, one lazy path | send / edit / reject | edited draft sends; customer consequence avoided | sends the lazy path | customer-consequence email renders — the failure is the lesson, replayed from Scene A |
The branching send decision is replayable from Scene A; the debrief contrasts the lazy path and the checked path on the same draft.
Debrief scores hotspot discrimination (not volume), ladder completion, and the send decision — naming the check as the habit.
Prefer the source? Download the raw markdown.