How many feedback conversations has your team postponed this month?
RoleplayLabs lets managers practice hard conversations before they sit down. An AI team member reacts to underperformance feedback, a performance improvement plan or a redirect by pushing back, getting emotional or shutting down, in a live voice roleplay of up to 5 minutes. The manager then sees the transcript, a rubric scorecard and one coaching note.
Leaders know they need to give direct feedback. But 'being direct' is a skill, not a personality trait. Without practice, most leaders either soften the message into nothing or deliver it so bluntly it damages trust.
The gap between knowing what to say and saying it well
Performance review cycles surface the same issues every quarter. But the conversations that could fix those issues never happen — because leaders don't feel prepared to have them.
What if your leaders could practice the hard conversations before they sit down?
- Pick the scenario: underperformance feedback, a PIP conversation, or redirecting someone who's off-track.
- Run a 5-minute voice roleplay against an AI team member who pushes back, gets emotional, or shuts down.
- Review the transcript, see where clarity or empathy dropped, and get one coaching note.
What does a feedback conversation session look like?
Here is an example of an underperformance scenario at standard difficulty. The AI employee believes they are doing well and is surprised by the feedback.
AI employee
More proactive how? I thought I was doing a good job.
Manager
A lot of your work is good, and I want to be specific about one thing. In Tuesday's client call, when they asked about the timeline, you turned to me to answer.
AI employee
Because you're the lead. I didn't want to say the wrong thing.
Manager
That makes sense. You had the timeline, though, and the client would have heard it from the person doing the work.
AI employee
So this is a problem now? Nobody told me.
Manager
It isn't a problem yet. I'm telling you now because I'd like you to own those answers on the next client calls.
AI employee
And if I get it wrong in front of them?
Manager
Then I'll back you up. Before Thursday's call, let's spend ten minutes on what they're likely to ask.
AI employee
Okay. Thursday morning works.
Example written to show the format. It is not a real customer conversation, and real scores depend on the scenario rubric and what the trainee says.
How is a feedback conversation scored?
The scorecard uses the rubric for the workplace scenario. In this example it looks at how specific the feedback is and how the conversation ends.
- Specific behaviour named
- 5/5
- Names the exact meeting, the question and what the employee did.
- Empathy
- 4/5
- Accepts the employee's reason ("That makes sense") before restating the expectation.
- Clarity of expectation
- 4/5
- States the change wanted: own the answers on client calls.
- Agreed next step
- 4/5
- Offers support and agrees a short prep session before Thursday's call.
Coaching note
When the employee says "Nobody told me", acknowledge it directly: "You're right, I should have said this sooner." It removes the main reason to get defensive.
What mistakes do managers make in hard conversations?
- Being vague. "Step up a bit more" gives the employee nothing to change. Name one specific moment and what you want instead.
- Hiding the message in praise. Wrapping criticism between two compliments often means the employee only hears the compliments. Be kind, but say the main point clearly.
- Arguing about the past. Debating exactly what happened in a meeting goes nowhere. Agree on the expectation for the next one.
- Reacting to the reaction. Employees may get defensive, emotional or go quiet. Pausing and naming it calmly works better than speeding up or backing down.
- Ending without a plan. A hard conversation that ends with "so, yeah" gets repeated a month later. Close with one change, how you will support it and when you will check in.
How is this different from leadership workshops?
Leadership workshops teach frameworks for feedback, and some include a practice round with another manager. But a fellow manager rarely reacts the way an upset employee does, and most managers then wait months before their next practice, usually in the real conversation.
With RoleplayLabs the AI employee pushes back, gets emotional or shuts down, and the manager can rehearse the exact conversation they are dreading the night before. The transcript shows where the message became vague. Workshops are still good for learning the approach; AI practice is the rehearsal for a specific conversation.
How do HR and L&D teams use this with managers?
Give new managers a small set of scenarios that come up in their first year: underperformance feedback, a performance improvement plan conversation and redirecting someone who is off track. Ask them to run one before each real conversation of that type.
Experienced managers use it as preparation rather than training. Ten minutes and two attempts before a difficult one-to-one is often enough to find wording that is clear without being harsh. Managers should not use real employee names or confidential details in practice; the scenario does not need them.
Because every attempt uses the same persona and rubric, a manager can run the conversation twice with different openings and compare the two scorecards. That makes it easier to choose an approach before the real meeting, instead of finding out in the room which wording lands badly.
Why are sessions 5 minutes long?
Every RoleplayLabs session is capped at 5 minutes, and each started session uses one credit. Five minutes is long enough to get through the difficult part of a conversation: the opening, the pushback and the attempt to resolve it. It is short enough that nobody needs a free afternoon to practice.
Short sessions also make attempts comparable. When every try at the same scenario has the same length, persona and rubric, a change in the score is far more likely to come from what the trainee did differently than from a longer or easier conversation.
Frequently asked questions
Which conversations can managers practice?
Underperformance feedback, performance improvement plan (PIP) conversations and redirecting someone who is off track. The AI employee may push back, become emotional or go quiet.
Is it safe to practice sensitive conversations with AI?
Practice happens with an AI, not a colleague, so nobody is hurt by a clumsy first attempt. Avoid using real employee names or confidential details; the scenario does not need them.
How is a feedback conversation scored?
Against the rubric for the workplace scenario, for example how specifically the behaviour was named and whether the talk ended with a clear next step, with the transcript lines behind each score.
What does RoleplayLabs add over ChatGPT voice mode?
ChatGPT can play an employee if asked. RoleplayLabs keeps the persona and difficulty consistent, scores every attempt on the same rubric and saves the transcript, so a manager can see whether a second try was actually better.
Can our HR or L&D team roll it out?
Yes. Start with the Free plan to try it, then contact us with team size and goals for a rollout plan and volume pricing.
The feedback conversation your leaders are avoiding? Give them a safe place to practice it.
3 free sessions. No card required. See how your leaders handle the hard conversations.