Digital Learning Practice
Sample 8-week journey: “Is it safe to put this in?”
Illustrative journey, not a contract template
Behaviour (break-room): You want to paste something into an AI tool and you are not sure you should.
Observable outcome: Before pasting, the person can classify the content (safe / needs stripping / must not go in), say why, and choose an alternative when the answer is no.
Audience example: Housing officers, caseworkers, or office staff with access to a chatbot or Copilot-style tool.
Entry: Team has done the free lesson path and/or AI at work, six weeks / One day, one team.
Horizon: Still able to decide correctly 60 days after the last taught session.
| Week | Phase | What happens | Learner action | Manager action |
|---|---|---|---|---|
| 0 | Discovery | Agree behaviour, rubric, privacy, baseline, delayed-check date | — | Sponsor signs measurement plan |
| 1 | Convert + baseline | Map objectives from existing teaching; baseline scenarios (no teaching in the test) | 10–15 min baseline judgements | Confirm real examples from the team’s work |
| 2 | Launch | Short refresh of the decision rules; first retrieval set with feedback | Classify 4–5 realistic pastes; explain one from memory | Observe one live decision that week if it occurs |
| 3 | Spaced practice | Gap; revisit; mix “tenant data”, “colleague email”, “public web text” (interleaving) | Answer before feedback; one summarise-from-memory card after a gap | Rubric: observed / not / N/A |
| 4 | Spaced practice | Harder variants; strip-and-retry patterns; when to use a local draft instead | Short scenarios + one “what would you remove?” task | Coaching cue on the most common miss |
| 5 | Spaced practice | Transfer tilt: new tools / new document types not in week 2 | New scenarios only | Spot-check two people |
| 6 | Light touch | Fewer items; critical objectives only; errors get a different explanation, not the same quiz again | 3–5 min | Team heatmap (patterns), not individual answer dumps |
| 7 | Delayed window opens | Unannounced-style delayed check (agreed in week 0) | Retention + new transfer scenario | Final observation pass |
| 8 | Readout | Plain report: retention, transfer, observations, guardrails, recommendation | Optional feedback on burden | Decide stop / revise / scale |
Example shape, not promised numbers
- Delayed correct classification rate vs baseline
- Transfer accuracy on unseen document types
- % of manager observations marked “observed safe decision”
- Time cost per person per week
- Opt-outs / complaints
Why not phishing for this sample? Phishing remains an excellent Pilot candidate (and the evidence page uses it to kill completion vanity). This sample matches the existing paid AI offers and the published free lesson “Is it safe to put this in?” so buyers see continuity.
Alternate sample (one-line): Same shape for “Tell a real message from a fake one” — practice must resemble the inbox decision; annual e-learning completion is not the endpoint.