Assignment
Use a simulated support workflow with weekly volume, baseline handling time, adoption rate, review time, and operating effort as editable assumptions.
Build sequence
- Define three hard gates and three measured improvement targets.
- Calculate low, expected, and high adoption scenarios.
- Add failure rework and maintenance effort to the time estimate.
- Specify the evidence and stakeholder decision required to expand or stop.
Acceptance checks
- The calculation can be reproduced from the stated assumptions.
- The scorecard does not equate model accuracy with business value.
- Simulated savings are never presented as measured customer outcomes.
Evidence to keep
Submit the working artifact or decision document described above, the inputs used, and the observed results for every acceptance check. Include one failed attempt and the change you made after investigating it. Label fixture-based outcomes separately from live-system measurements. These artifacts become part of your final portfolio review.
Guided workshop: define a decision the pilot can answer
Continue with Northstar Support. The pilot asks whether a draft assistant reduces handling time without worsening answer quality or access control. It does not ask whether a polished demo impresses the sponsor.
Use this fictional baseline: 100 eligible tickets per week, eight minutes of handling time each. The pilot estimates five minutes with assisted drafting, including review. At 60% adoption, expected saved time is 100 × (8 − 5) × 0.60 = 180 minutes, or three hours per week. This is a forecast until measured.
Sensitivity table: calculate saved time at adoption rates 30%, 60%, and 90%, and assisted handling times of four, five, and seven minutes. Include training, review, integration, and operating effort when discussing value. Time saved is not automatically cash saved or realized business value.
Define the scorecard before running the pilot:
| Dimension | Exercise gate | Evidence |
|---|---|---|
| Access | Zero forbidden evidence exposed | Tenant matrix tests |
| Grounding | At least 18 of 20 supported cases correct | Reviewed case results |
| Unsupported requests | All five routed for review | Frozen negative cases |
| Usefulness | Median handling time improves | Paired task observations |
| Operations | Operator can disable drafting | Handoff rehearsal |
These are teaching targets for a small simulation, not statistically established production guarantees. Choose actual thresholds with the customer according to harm, frequency, and acceptable fallback.
Worked decision: a pilot meets average time savings but fails one cross-tenant test. Stop expansion and fix authorization. Another pilot meets quality gates but operators do not use it; investigate workflow placement, trust, and training before adding features.
Exercise: distinguish a proof of concept, pilot, and production service. The proof of concept tests technical feasibility. The pilot tests a bounded operational hypothesis with named users and support. Production requires sustained ownership, access controls, monitoring, recovery, and a rollout decision.
Assessment: submit the baseline, assumptions, sample plan, scorecard, sensitivity calculations, and a signed-off simulated decision owner. State what result means expand, revise, or stop. An undefined success criterion allows every outcome to be described as a success and prevents an honest customer recommendation.