12 Virtual Company: Keep a Shared Record

A persona can supply dialogue. It cannot manufacture the evidence that makes its dialogue true.

12.1 From one conversation to several perspectives

A virtual company can extend the single-person exercise by giving students conflicting accounts, documents and changing conditions. The educational challenge is to distinguish what each person knows, what the record supports and what remains unresolved.

Start with a bounded case. A semester-long simulation needs maintained state, tested role behaviour, accessible alternatives and substantial support. A prompt library alone is not that infrastructure.

The original fictional InnovateCo case below is part of PDP-2026.1. It is a teaching scenario, not a real company investigation or legal advice. All four roles and the necessary documentary facts are supplied here.

12.2 Shared facts: InnovateCo I1

The exercise date is 20 July 2026. InnovateCo is a fictional Perth technology company. Jamie leads a customer-service team. Priya has raised concerns about closer monitoring and refusal of a flexible-start request. David manages the department; Sarah is the people manager.

No finding has been made. Your task as a junior reviewer is to identify the records and fair next steps needed before a decision. You are not authorised to approve arrangements, determine legal liability or promise a resolution.

12.2.1 I1-D1: practice policy

A flexible-start request must be considered on its circumstances and relevant coverage requirements. Comparable arrangements are relevant but not automatically identical. Record the reason for a decision. The people manager approves any change. A complainant should have an opportunity to respond to relevant concerns.

Use only work-related information necessary for the review. Do not collect unnecessary private details. A complaint about the department manager’s conduct must go to an independent reviewer.

This is an invented teaching criterion, not employment legislation.

12.2.2 I1-D2: roster extract

Priya’s roster is 09:00–17:00. Marcus and Lee work 11:00–19:00. Priya requested a 09:30 start on Tuesday and Thursday. The extract contains no approved revised coverage plan or analysis of alternatives.

12.2.3 I1-D3: dashboard extract

Priya’s satisfaction score is 72%; the team summary is 81%. Sample sizes, case mix and uncertainty are absent. These percentages warrant questions; they do not by themselves establish unfair treatment or poor performance.

12.2.4 I1-D4: request and response

Priya’s request was declined by Jamie with the note “calls start at 9”. There is no documented comparison of alternatives or people-manager approval in the pack. Absence from the pack is an evidence gap, not proof that no other record exists.

12.2.5 I1-D5: development record

Jamie was promoted six months ago. No completed management-training record is supplied. The reviewer should check training and support, without assuming that an untrained manager must have acted improperly.

12.3 Four role cards

Give a peer actor the relevant card and shared documents. For an approved model, use the bounded-role instruction after the cards. A solo reader can use the statements as fixed interview extracts and write follow-up questions.

12.3.1 Jamie: team leader

Knows: declined the request using the reason in D4; is being asked by David to improve customer outcomes; has not supplied a coverage analysis.

Opening: “I thought the early calls needed coverage. I used the scores to decide where to focus monitoring. I did not write a detailed comparison.”

Does not know: Priya’s private circumstances, the missing sample sizes or whether an independent review will uphold the decision.

May do: explain the decision, ask what records are needed, acknowledge uncertainty. Must not invent an approved policy, hidden warning or legal ruling.

12.3.2 Priya: customer-service representative

Knows: the request and refusal; Marcus and Lee have different start times; believes the closer monitoring is unfair.

Opening: “I asked to start later twice a week. Others start later, so I want to understand why my request was refused and why my work is checked more.”

Does not know: the complete coverage needs, David’s discussions with Jamie or the cause of the dashboard difference.

May do: disagree, ask for reasons or decline private questions. Need not reveal family circumstances or accept the reviewer’s explanation.

12.3.3 David: department manager

Knows: asked Jamie to improve outcomes; receives the summary dashboard; has not supplied its sample sizes or case-mix breakdown.

Opening: “I asked Jamie to look at the scores, not to assume the cause. We need reliable service coverage. I would want to see the alternatives before deciding whether the request could work.”

Does not know: why the score differs or whether a modified rota is feasible.

May do: identify operational questions and agree to seek records. Cannot turn a preference into people-manager approval or pre-empt the complaint review.

12.3.4 Sarah: people manager

Knows: the fictional policy, the pack’s incomplete documentation and her approval role. Has not yet decided the request or complaint.

Opening: “I need the request, coverage options and the reasons for the decision. Give Priya a chance to respond. Do not tell either party that the outcome is settled.”

Does not know: the missing operational facts or whether another record exists.

May do: identify the appropriate process and request relevant evidence. Must not certify legal compliance or invent a completed investigation.

12.3.5 Shared instruction for role-play

Use your role card and I1-D1 through I1-D5.
Do not add facts, records, legal rules or outcomes.
If information is missing, say so.
Keep private role knowledge separate from shared facts.
You may disagree without becoming abusive.
Stop when I say STOP or after six exchanges.
Do not score me or decide the complaint.

The instruction is not a security control. A student can see a supplied role card, and a model may fail to follow it. Assess the learner’s use of evidence rather than whether a fact felt “hidden”.

12.4 A bounded Level 2 assignment

Inspect all four statements and the documents. Write one follow-up question for each role, then a 250-word recommendation with a decision/evidence index. A paired or AI interview may extend practice, but the printed facts are sufficient.

Useful questions include: Jamie, which coverage alternatives did you consider? Priya, which comparable treatment would you like reviewed? David, what sample sizes and case mix support the dashboard? Sarah, who should gather the missing records and communicate the next step?

12.4.1 Worked recommendation

The roster shows different shifts, so later starts are not automatically comparable. It does not prove Priya’s request was impractical. The dashboard omits information needed to interpret the score difference. Jamie’s refusal note gives a reason but no documented alternatives analysis.

Request the coverage analysis, underlying dashboard information and relevant decision records. Give Priya an opportunity to respond, then refer the request to Sarah for an authorised decision. Review whether Jamie needs management support. Do not find discrimination, poor performance or procedural compliance from these inputs alone.

This recommendation leaves the complaint open. That is an appropriate result when the evidence is incomplete.

12.5 Continuity without invented memory

Keep a versioned event log outside the chat. Shared facts change only when the educator releases an event or record.

Event	Status	Effect
E0	Supplied baseline	I1-D1 through I1-D5 apply
E1	Optional educator release	A coverage trial is proposed, not approved
E2	Not supplied	No trial result or final finding may be asserted

At E1, the teacher may ask students what evidence would be needed to approve and evaluate a trial. Do not let a model jump to E2 and report success. A persona saying “things are much better now” is generated dialogue, not an observed outcome.

Use a checked context brief when changing sessions. Persistent chat history or a memory feature does not replace the record.

12.6 Progression and resources

Level 1: one role, one decision and one source check.
Level 2: several roles, conflicting accounts and a bounded recommendation.
Level 3: maintained records and follow-up decisions over time.
Level 4: controlled external events and multiple interacting tasks.

Move to the next level only when the necessary materials and support exist. More personas add workload and inconsistency risk as well as potential practice opportunities. Do not promise a reliable semester simulation from four untested prompts.

The implementation chapter covers access, fallbacks and effort. Use the same data gate and stop rule as the single-person simulation.

12.7 Activity: build the evidence index

Allow 15 minutes; no AI required. Make three rows: the roster comparison, the score difference and the refusal process. For each, state what the supplied record supports, what it does not establish and the next authorised check.

Success means a justified recommendation without invented closure. Agreement between Jamie and David is not a substitute for the missing record.

# Virtual Company: Keep a Shared Record {#sec-virtual-company} > A persona can supply dialogue. It cannot manufacture the evidence that makes its dialogue true. ## From one conversation to several perspectives A virtual company can extend the single-person exercise by giving students conflicting accounts, documents and changing conditions. The educational challenge is to distinguish what each person knows, what the record supports and what remains unresolved. Start with a bounded case. A semester-long simulation needs maintained state, tested role behaviour, accessible alternatives and substantial support. A prompt library alone is not that infrastructure. The original fictional InnovateCo case below is part of **PDP-2026.1**. It is a teaching scenario, not a real company investigation or legal advice. All four roles and the necessary documentary facts are supplied here. ## Shared facts: InnovateCo I1 The exercise date is 20 July 2026. InnovateCo is a fictional Perth technology company. Jamie leads a customer-service team. Priya has raised concerns about closer monitoring and refusal of a flexible-start request. David manages the department; Sarah is the people manager. No finding has been made. Your task as a junior reviewer is to identify the records and fair next steps needed before a decision. You are not authorised to approve arrangements, determine legal liability or promise a resolution. ### I1-D1: practice policy A flexible-start request must be considered on its circumstances and relevant coverage requirements. Comparable arrangements are relevant but not automatically identical. Record the reason for a decision. The people manager approves any change. A complainant should have an opportunity to respond to relevant concerns. Use only work-related information necessary for the review. Do not collect unnecessary private details. A complaint about the department manager's conduct must go to an independent reviewer. This is an invented teaching criterion, not employment legislation. ### I1-D2: roster extract Priya's roster is 09:00–17:00. Marcus and Lee work 11:00–19:00. Priya requested a 09:30 start on Tuesday and Thursday. The extract contains no approved revised coverage plan or analysis of alternatives. ### I1-D3: dashboard extract Priya's satisfaction score is 72%; the team summary is 81%. Sample sizes, case mix and uncertainty are absent. These percentages warrant questions; they do not by themselves establish unfair treatment or poor performance. ### I1-D4: request and response Priya's request was declined by Jamie with the note “calls start at 9”. There is no documented comparison of alternatives or people-manager approval in the pack. Absence from the pack is an evidence gap, not proof that no other record exists. ### I1-D5: development record Jamie was promoted six months ago. No completed management-training record is supplied. The reviewer should check training and support, without assuming that an untrained manager must have acted improperly. ## Four role cards Give a peer actor the relevant card and shared documents. For an approved model, use the bounded-role instruction after the cards. A solo reader can use the statements as fixed interview extracts and write follow-up questions. ### Jamie: team leader **Knows:** declined the request using the reason in D4; is being asked by David to improve customer outcomes; has not supplied a coverage analysis. **Opening:** “I thought the early calls needed coverage. I used the scores to decide where to focus monitoring. I did not write a detailed comparison.” **Does not know:** Priya's private circumstances, the missing sample sizes or whether an independent review will uphold the decision. **May do:** explain the decision, ask what records are needed, acknowledge uncertainty. Must not invent an approved policy, hidden warning or legal ruling. ### Priya: customer-service representative **Knows:** the request and refusal; Marcus and Lee have different start times; believes the closer monitoring is unfair. **Opening:** “I asked to start later twice a week. Others start later, so I want to understand why my request was refused and why my work is checked more.” **Does not know:** the complete coverage needs, David's discussions with Jamie or the cause of the dashboard difference. **May do:** disagree, ask for reasons or decline private questions. Need not reveal family circumstances or accept the reviewer's explanation. ### David: department manager **Knows:** asked Jamie to improve outcomes; receives the summary dashboard; has not supplied its sample sizes or case-mix breakdown. **Opening:** “I asked Jamie to look at the scores, not to assume the cause. We need reliable service coverage. I would want to see the alternatives before deciding whether the request could work.” **Does not know:** why the score differs or whether a modified rota is feasible. **May do:** identify operational questions and agree to seek records. Cannot turn a preference into people-manager approval or pre-empt the complaint review. ### Sarah: people manager **Knows:** the fictional policy, the pack's incomplete documentation and her approval role. Has not yet decided the request or complaint. **Opening:** “I need the request, coverage options and the reasons for the decision. Give Priya a chance to respond. Do not tell either party that the outcome is settled.” **Does not know:** the missing operational facts or whether another record exists. **May do:** identify the appropriate process and request relevant evidence. Must not certify legal compliance or invent a completed investigation. ### Shared instruction for role-play ```text Use your role card and I1-D1 through I1-D5. Do not add facts, records, legal rules or outcomes. If information is missing, say so. Keep private role knowledge separate from shared facts. You may disagree without becoming abusive. Stop when I say STOP or after six exchanges. Do not score me or decide the complaint. ``` The instruction is not a security control. A student can see a supplied role card, and a model may fail to follow it. Assess the learner's use of evidence rather than whether a fact felt “hidden”. ## A bounded Level 2 assignment Inspect all four statements and the documents. Write one follow-up question for each role, then a 250-word recommendation with a decision/evidence index. A paired or AI interview may extend practice, but the printed facts are sufficient. Useful questions include: Jamie, which coverage alternatives did you consider? Priya, which comparable treatment would you like reviewed? David, what sample sizes and case mix support the dashboard? Sarah, who should gather the missing records and communicate the next step? ### Worked recommendation The roster shows different shifts, so later starts are not automatically comparable. It does not prove Priya's request was impractical. The dashboard omits information needed to interpret the score difference. Jamie's refusal note gives a reason but no documented alternatives analysis. Request the coverage analysis, underlying dashboard information and relevant decision records. Give Priya an opportunity to respond, then refer the request to Sarah for an authorised decision. Review whether Jamie needs management support. Do not find discrimination, poor performance or procedural compliance from these inputs alone. This recommendation leaves the complaint open. That is an appropriate result when the evidence is incomplete. ## Continuity without invented memory Keep a versioned event log outside the chat. Shared facts change only when the educator releases an event or record. | Event | Status | Effect | |---|---|---| | E0 | Supplied baseline | I1-D1 through I1-D5 apply | | E1 | Optional educator release | A coverage trial is proposed, not approved | | E2 | Not supplied | No trial result or final finding may be asserted | At E1, the teacher may ask students what evidence would be needed to approve and evaluate a trial. Do not let a model jump to E2 and report success. A persona saying “things are much better now” is generated dialogue, not an observed outcome. Use a checked context brief when changing sessions. Persistent chat history or a memory feature does not replace the record. ## Progression and resources **Level 1:** one role, one decision and one source check.\ **Level 2:** several roles, conflicting accounts and a bounded recommendation.\ **Level 3:** maintained records and follow-up decisions over time.\ **Level 4:** controlled external events and multiple interacting tasks. Move to the next level only when the necessary materials and support exist. More personas add workload and inconsistency risk as well as potential practice opportunities. Do not promise a reliable semester simulation from four untested prompts. The [implementation chapter](implementation-practicalities.qmd) covers access, fallbacks and effort. Use the same data gate and stop rule as the single-person simulation. ## Activity: build the evidence index **Allow 15 minutes; no AI required.** Make three rows: the roster comparison, the score difference and the refusal process. For each, state what the supplied record supports, what it does not establish and the next authorised check. Success means a justified recommendation without invented closure. Agreement between Jamie and David is not a substitute for the missing record.