The course outcome is not a line count or a claim that “AI built it”. It is a small solution with clear boundaries, a working main path, verified errors, and instructions another person can follow.
Capstone handoff
- Brief: audience, journey, scope, and acceptance criteria.
- Environment decision: why it fits and how to recover the source.
- Agent tasks: persistent context plus small reviewable changes.
- Working journey: one user path from input to result.
- Data and access: fields, storage, roles, allowed and denied actions.
- Verification: scenarios, defects found, and retests.
- Release: preview, version notes, limitations, and rollback.
- Five-minute demo: problem, journey, failure/limit, and next step.
Score the result
Score each item 0–2: 0 means missing or uncheckable; 1 means present but weakly evidenced; 2 means another person can verify and repeat it.
| Criterion | Score |
|---|---|
| Problem and acceptance criteria are specific | /2 |
| Version 1 is bounded | /2 |
| Main journey works and reports truthfully | /2 |
| Errors and phone layout were checked | /2 |
| Data access and secrets were reviewed | /2 |
| Setup instructions and a saved version exist | /2 |
| Release can be checked and rolled back | /2 |
| Limitations are stated without exaggeration | /2 |
| Total | /16 |
Aim for at least 12/16, with no zero for checks, data, or limitations. This is a learning rubric, not a certification or proof that real user data is safe.
A five-minute demo
- 0:00–0:40: user and the manual step you address.
- 0:40–2:30: the main journey, without skipping steps.
- 2:30–3:30: one failure and its safe fallback.
- 3:30–4:20: tests and limitations.
- 4:20–5:00: the next thing to validate with a real process owner.
Practice
- Assemble the capstone handoff checklist.
- Ask another person to score it.
- Fix one gap rather than hiding it in a presentation.
- Record your demo and verify every claim against evidence.
Final artifact: prototype/preview, project folder, README, test table, release notes, and short demo.
What you can accurately say
“I built a learning web prototype for consultation requests with an AI tool; wrote acceptance criteria; checked error states and the mobile layout. Real data integration and a customer pilot were not performed.” Name the tools and your own contribution. Do not call a learning prototype a production implementation or claim unmeasured business impact.