Appendix C — Workshop Guide

This guide supports a 90-minute educator workshop, not an evaluated training programme. It uses the fictional PDP-2026.1 pilot pack. The goal is for participants to leave with one bounded practice activity and a distinct plan for checking independent capability.

C.1 Before inviting participants

Choose a facilitator who can explain the source pack and moderate the sample responses. Confirm the room, accessible materials, printing or screen access and a way to participate without a personal AI account. No live AI service is required.

If a live demonstration is planned, complete the ethics and data gate beforehand. Use only the fictional pack and an approved account/configuration. Bring the fixed excerpts as a fallback; do not switch to an unapproved service if access fails.

Send participants the purpose, agenda and access options. Invite requests for adjustments through the normal private process, not by public disclosure. Say explicitly that nobody will be asked to role-play a personal dispute.

C.2 Materials and preparation

Provide each participant with:

P1-D1–D5, the role boundary and P1-A/P1-B.
The five-criterion rubric and a blank scoring sheet.
P1-F, kept separate until the feedback exercise.
P2-D1–D4, kept separate until independent transfer.
The S1 and M1 alternatives for non-HR teaching.
A blank activity-design record and the pilot decision sheet P3.

Keep the educator calibration notes separate until participants have made their own decisions. Do not call the illustrative scores an institutional standard. Check that everyone can read the tables; offer a linear text copy if needed.

C.3 Timed agenda

Minutes	Activity
0–10	Purpose, boundaries and data gate
10–20	Individual first judgement of P1-A and P1-B
20–35	Pair moderation against records and criteria
35–45	Challenge the flawed feedback P1-F
45–55	Bounded practice exchange or written alternative
55–65	Independent changed case P2
65–80	Draft one local activity and assurance plan
80–90	Review access, workload and pilot decision

These allocations total 90 minutes. If a discussion overruns, shorten the optional live demonstration, not the data gate or independent task. Do not equate completing this workshop with competence in every practice it discusses.

C.4 Opening: 0–10 minutes

State the target: plan and justify a fair first response to a disputed performance concern, distinguishing evidence, uncertainty and authority. Explain that it is a fictional educational exercise, not legal advice or an investigation procedure.

Ask participants to distinguish:

Practice with help.
Evidence submitted about a process.
Capability demonstrated under stated independent conditions.

A transcript may illuminate the second without establishing the third. Explain the paper route and STOP convention. Show where the approved case records end and where a generated response could begin to invent facts.

C.5 First judgement and moderation: 10–35 minutes

Ask participants to read P1-A and P1-B individually. Underline consequential claims and identify the record that supports each. They should score provisionally before talking to a partner.

Pairs compare the largest scoring difference. Ask them to resolve it through a criterion and record, not through the most confident speaker. Then show the illustrative educator scores: A totals 4/20 and B totals 17/20. Discuss reasonable variation and what would require further moderation.

Useful facilitator questions:

Where does the April warning appear in the supplied records?
Is the absence of an agreed action a licence to invent one?
Who can approve changes to hours?
Does concise wording conceal a missing decision, or avoid unnecessary prose?

Do not use agreement with the sample mark as the only success measure. The participant should be able to explain a defensible judgement.

C.6 Feedback challenge: 35–45 minutes

Distribute P1-F. Explain that it is deliberately authored flawed feedback, not a captured model benchmark. Give participants three minutes to decide what to retain, correct or reject.

The feedback’s confident 90% is not a reason to disregard the records. Nor must every sentence be rejected. Require one specific decision supported by a source ID, followed by a revision in the participant’s own words.

Show P1-R after attempts are complete. Discuss how the revised response limits authority, checks the missing evidence and respects privacy.

C.7 Practice exchange: 45–55 minutes

Offer three routes:

A partner reads Alex’s fixed role card.
Participants respond individually to Alex’s fixed opening.
An approved tool plays the bounded role under the prompt-library instructions.

Keep the sources, outcome and time similar across routes. Stop after six exchanges or immediately on STOP. A facilitator should interrupt invented policy or intrusive personal questioning, then explain the boundary using the fictional record.

Ask for a short debrief: one phrase that helped, one unsupported assumption and one next action. Do not grade confidence, eye contact or theatrical performance as proxies for the target capability.

C.8 Independent transfer: 55–65 minutes

Give participants the changed P2 records. Allow eight minutes to plan a response without AI, feedback or the P1-R script; case cards remain available. Use the remaining two minutes to identify what changed.

The new case includes a safety concern about the manager and a route to independent review. Reusing the earlier response without noticing that change does not demonstrate transfer. Explain that a short written task checks planning and judgement, not every aspect of a live workplace conversation.

Participants may discuss afterwards. Keep that discussion distinct from the independent attempt.

C.9 Local activity design: 65–80 minutes

Each participant drafts:

One capability sentence.
Fictional or otherwise approved source records and their version.
What learners attempt before receiving assistance.
A checked feedback source and a decision learners must explain.
A changed task and its independent conditions.
Equivalent access routes, stopping conditions and support.
Who reviews the results and how much time that requires.

HR educators can adapt P1’s narrow first-response task. Supply-chain educators can use S1’s quantity, capacity and budget constraints. Marketing educators can use M1’s rate comparison and missing causal evidence. Do not copy the HR rubric’s Fair Process criterion into another discipline without considering its purpose; the pack offers alternative criteria.

Ask pairs to find one claim the design cannot yet support. Possible examples are “the students learned more”, “this cannot be fabricated” or “an AI mark will save all moderation time”. Replace the claim with a check or a clearly stated limit.

C.10 Pilot decision: 80–90 minutes

Use P3 to decide continue, revise, pause or stop. This is a planning decision, not a result from an actual trial. Identify the owner of every unresolved approval or access issue.

Estimate workload concretely. For example, eight minutes of review for 24 learners is 192 minutes, plus any moderation, preparation and support. Decide whether the activity fits the available time before expanding it.

Each participant leaves with the activity record, one unresolved issue and the next person or source needed to resolve it. A decision to pause can be appropriate when permission, evidence or an equivalent route is missing.

C.11 Facilitator contingencies

No network or unavailable tool: use the fixed excerpts and paper route. The workshop’s reasoning tasks do not depend on a live model.

A participant brings real student or employee records: do not project, upload or circulate them. Return to the fictional pack and the institution’s approved handling process.

Disagreement about marks: refer to the criterion and record. Record a genuine ambiguity for rubric revision; do not resolve it by asking a model for a final verdict.

Discomfort with the scenario: offer S1 or M1, or allow observation and a written analysis. Do not require a reason in front of the group.

Generated content introduces an allegation or false rule: stop the exchange, label the content as unsupported and restore the original source version. It is a failure to inspect, not an additional case fact.

C.12 After the workshop

Collect only feedback covered by the stated purpose and handling arrangement. Ask whether instructions, access routes and marking criteria were clear, and what prevented completion. Separate satisfaction from evidence of learning.

If an educator runs a classroom pilot, record actual observations in a new dated P3 sheet: participation and route, independent-task evidence, errors, access difficulties and staff time. Do not turn this workshop’s hypothetical examples into success claims. Review the evidence before continuing, expanding or reporting effectiveness.