Download the complete part-08 practice pack (ZIP). Extract it and preserve its folders. All organizations, records, and results in this teaching pack are fictional.
Use these templates to plan work with the people involved. All accompanying examples are fictional; no training, consultation, applicant assessment, or pilot has been conducted. Proposed role assignments and policies need an actual decision before they govern a workplace.
Role-based learning plan
| Field | Entry to complete |
|---|---|
| Business task | A specific output the learner needs to produce |
| Learner role | Responsibility and actual authority |
| Learning objective | Observable task, source conditions, and review criteria |
| Prerequisites | Access, source knowledge, and basic skills required |
| Source packet | Approved documents and versions |
| Practice case | Ordinary example plus a relevant exception |
| Critical errors | Mistakes that require stopping or returning to supervised practice |
| Feedback | Specific correction and another suitable attempt |
| Access and support | Usable format, working time, assistance, and confidential request route |
| Follow-up | Owner, date, and evidence of application to work |
Keep attendance separate from competence. Keep an exercise result separate from a personnel decision. Do not turn the proposed rubric into a company-wide certification or hiring test without the relevant review.
Trainer's sixty-minute session outline
- Ten minutes: Explain the task, allowed sources, and limits on actions. Ask what makes this task difficult in practice.
- Ten minutes: Demonstrate checking the current policy and identifying a missing fact. Show the original document beside the draft.
- Twenty minutes: Let the learner attempt a new case. Offer the access support needed without supplying the substantive answer.
- Fifteen minutes: Review against the rubric, explain the specific gap, and let the learner correct it.
- Five minutes: Agree the next practice or later work-sample review and its owner.
Adapt the time to the learner and task. The suggested schedule totals sixty minutes; it is not an assessment time limit or evidence that training occurred.
Use the learner-task column in practice/training-cases.csv to prepare learner materials. Keep the expected-review and critical-issue columns as trainer notes until the review step. A trainer should check the answer key against the preserved Mesa sources before running the exercise.
Consequential employment-use review record
Proposed use and employment decision affected:
Jurisdiction and business/job context:
Actual job duties and required evidence:
Data used and why it is necessary:
Validity and relevant evaluation evidence:
Access and accommodation process:
Potential discrimination and privacy issues:
Notice, correction, and challenge process:
Qualified reviewers and decision owner:
Decision, conditions, and review date:
This record organizes questions. It is not legal clearance or authorization to rank applicants, infer personal traits, or make employment decisions. Drafting a job description from supplied duties does not answer these questions for a candidate-scoring system.
Pilot agreement
Specify the problem, task, eligible and excluded cases, proposed period, responsible people, permitted data, and review requirements. Record the staff input that changed the proposal. Leave unresolved decisions visible.
Add the measures and their denominators before looking at outcomes. Include full effort, quality, exceptions, and staff experience. Define pause conditions, a manual fallback, and who can approve a restart or expansion.
Explain what logs are collected, who can access them, how long they are kept, and their permitted use. Keep operational improvement records separate from any proposed consequential employment use. Do not promise anonymity, confidentiality, employment protections, or support arrangements that the actual process cannot provide.
Feedback and response record
| Field | Meaning |
|---|---|
| Concern ID | Stable reference |
| Concern and evidence | The actual issue, preserving uncertainty |
| Proposed action | What might address it |
| Responsible person | Confirmed assignment, not merely a suggested role |
| Implementation status | What was actually changed |
| Resolution evidence | How the change addresses the concern |
| Confirmation and review date | Who reviewed the resolution and what remains open |
Drafted responses do not close concerns. In the constructed feedback register, no implementation or confirmation is recorded. Do not convert four example concerns into a survey prevalence estimate.
Decision record
Decision question, owner, and deadline:
Purpose and consequences of waiting:
Facts and source versions:
Assumptions to verify:
Preferences and who owns them:
Unknowns:
Options, including the current process:
Essential conditions and excluded options:
Specified cash costs and omitted resource costs:
Comparison and sensitivity:
Disagreement and evidence that could change the choice:
Proposed action and scope:
Actual approval, when recorded:
Review date and outcomes, when available:
A score is a summary of chosen inputs. It does not supply missing evidence or approve an action. Keep a failed essential condition outside the ranked shortlist. Record whether a recommendation is to investigate, test, buy, deploy, or expand; those actions require different evidence and authority.
Practice sequence and files
Read practice/Practice-Sources.md, then the preserved Mesa documents in practice/mesa-sources/. Use the six training cases and five hiring-boundary cases to practice source use and scope decisions. Observed fields remain blank until an actual evaluation is performed.
Use constructed-training-cohort.csv and training-denominators.csv to explain the difference between attendance, completion, and demonstration. No baseline learning or follow-up result is supplied.
Trace constructed-adoption-log.csv and adoption-assumptions.json into adoption-calculations.csv. Check why uptake is 60% of eligible requests, why first-review acceptance is about 83.3%, and why full effort gives only eight modeled minutes of capacity difference.
Inspect decision-inputs.json, decision-comparison.csv, and decision-record-draft.json. Explain why C is not scored and why the ranking of A and B reverses when B's effort score changes. The draft recommendation is to investigate, not to deploy.
Run python3 practice/verify_examples.py from the Part 8 directory to reproduce the calculation CSVs and Example-Verification.json. The script uses Python's standard library and checks arithmetic and constructed-record consistency. It does not evaluate a person, an AI model, workplace fairness, legal compliance, or a real operating system.