What you'll learn
- Build a rehearsal brief from facts, agreed preferences and unanswered questions.
- Practise receiving different accounts without assigning a personality to your colleague.
- Use Cooperly context in approved AI preparation, then confirm the next step with the person.
How should AI roleplay for managers use team context?
A manager can enter a practice chat with a conclusion already attached: 'Play my colleague, who avoids difficult discussions.' Every generated reply then arrives inside that story. The exercise may help you rehearse confronting avoidance before you have established that avoidance occurred.
Begin with an observable work event and a question you need answered. A cancelled review invitation is an event. 'They do not want scrutiny' is an explanation to investigate. The colleague may have moved the meeting, followed a pause you requested or misunderstood its purpose.
In the exploratory 2025 CommCoach study, managers described the value they saw in relevant practice scenarios, feedback and human support. The research examined their perceptions of a prototype; it did not establish better real conversations or lasting skill gains. Treat the exercise here as a suggested way to prepare, then check it against the actual conversation.
Build a brief that leaves room for correction
Consider a hypothetical design review. Its calendar invitation was cancelled, and the manager cannot find a replacement. The colleague has previously asked to receive the specific example before a work discussion. That request gives the manager a useful preparation step: send the invitation and explain what needs clarifying.
| Context | What to supply |
|---|---|
| Checked event | The design-review invitation was cancelled. I cannot find a replacement in the calendar. |
| My purpose | Find out what changed and agree when the design will be reviewed before the customer discussion. |
| Confirmed preference | My colleague has asked to see the specific example before a work discussion. I will send it first. |
| Still unknown | Whether another time was agreed, whether I requested a pause, and what the colleague understood the review to cover. |
| What I will practise | Ask for their account, acknowledge a correction and clarify a feasible next step. |
Use the preference only if it is current, relevant and allowed in the AI tool you plan to use under your team's rules. If it is unavailable, leave it out. You can still practise with a synthetic scenario. A short approved statement is enough; the tool does not need a full profile or a history of private messages.
The 2026 ASPECT preprint examined AI-inferred profiles from workplace communication in one organisation. Participants identified errors involving context, tone and who had said something. The study concerned initial judgments of profiles and hypothetical responses, not this rehearsal method. It gives a reason to check a profile's description with the person rather than treat it as a complete account of them.
Bring Cooperly context into your chosen AI tool
Cooperly's People Context Layer connects working profiles, team signals, goals, facts and team memory to approved AI surfaces. The context available depends on the user, team and integration permissions. For the design-review example, choose only the relevant work context and permitted preference; leave unrelated records out.
A Coop Profile can start a discussion about working preferences. Check what applies with the colleague. A team-level anonymous signal is not evidence of this person's reason for cancelling a meeting, and a profile description does not establish their current feelings.
The hypothetical brief above can be supplied as synthetic text; it is not a report from a connected customer account. Before using real context, follow your team's approved AI-use rules. Permission to read something does not automatically make every onward use appropriate. Keep sensitive material outside the practice request.
What to practise in AI roleplay for managers
Try an opening aloud or in writing. 'Why are you avoiding the review?' assumes a motive. A more useful opening might be: 'The review invitation was cancelled, and I cannot find the replacement. What changed? We need an agreed review point before the customer discussion.'
- 01
Practise a correction to your account
A fictional reply might say a replacement was agreed in a thread you missed. Practise asking to check that message and acknowledging the update, rather than arguing from the calendar alone.
- 02
Practise discovering your own contribution
Another possible reply might say you asked for a pause while customer feedback arrived. Check that instruction. If it is yours, clarify what is now ready for review instead of continuing to accuse the colleague of delay.
- 03
Practise a disagreement about the next step
The counterpart might say the review question or attendees are unclear. Practise explaining the decision needed, asking what input is missing and agreeing a realistic next step. Do not demand agreement merely to finish the exercise.
These are illustrative possibilities, not predictions or a required number of repetitions. Keep each simulated account separate from the facts in your brief. If the model invents a private reason or treats the profile as an explanation, stop and remove that assumption before continuing.
One possible revised response is: 'I may have missed the update. Can we check where the new time was agreed? Then let's make sure the people who need to review can find it.' The practice concerns whether you can adjust your question while keeping the work requirement clear.
The 2026 Science research on AI sycophancy found that overly affirming AI advice affected participants' immediate judgments and willingness to repair interpersonal conflict. It did not measure later relationship repair or manager roleplay. When feedback tells you that you handled everything perfectly, ask which words support that assessment and examine them yourself.
Take the questions back to the person
Send the example ahead if that is the agreed preference, then have the conversation. The colleague may offer an account none of the practice branches included. Hear it, check what needs checking and revise the preparation. A generated line from the exercise is not evidence that they said or intended anything.
For the hypothetical review, the useful outcome is an agreement people can act on: what needs reviewing, who will participate and when. If a pause was genuinely requested, make the restart condition clear. If a replacement already exists, make its location visible to the affected people. Resolve what actually happened before documenting a cause.
If the conversation changes a working agreement, update the appropriate record through the authorised human process. Keep the practice transcript and invented replies out of team facts. The shared AI context guide explains how to keep a changed agreement current; the read-only connection does not write the new agreement for you.
When AI roleplay for managers needs human support
If you have rehearsed repeatedly but still have no genuine question for the colleague, stop generating more replies. Ask a trusted, appropriately informed person to challenge your opening. Keep their access to the case proportionate, and do not let practice become a reason to postpone a necessary conversation.
For formal employment decisions, complaints, safety issues or health concerns, use the relevant qualified support and procedure. Practising wording can be one preparation step; it does not establish the facts or decide the process.
Start with one ordinary conversation. Prepare a small factual brief, practise receiving a different account and take your questions to the person. The next thing to check is whether your real conversation produced a clearer, mutually understood next step.