Level 16 min
Open
Kenji is frustrated but wants this fixed. He admits the public revert as soon as you ask what happened after he got back.
- One-on-one
- Owns his part when asked
Leadership roleplay scenarioConflict resolution role play scenario: “She merged my code while I was out”
Two engineers on your team have stopped working together, and the release is slipping. Practice hearing one side without taking it, naming the cost to the team and settling on a plan with Kenji, an AI backend engineer who expects to be blamed.
KenjiAI backend engineer
Mood · AI character
If this is about Marta, ask her why she merged my code while I was on leave.
I’ll talk to Marta too. Right now I want your side. What did you find when you got back?
Half my retry logic rewritten. Nobody asked me. She wants the payments service, that’s obvious.
Scored live
- Stays neutral
- Separates facts from assumptions
- Names the impact on the team
- Gets him to own his part
- Agrees on a plan
AI coachYou asked for his account before judging anyone. Next, check what he assumes about her.
The situation
Two strong engineers. One merge. Three weeks of silence.
Context
You manage a team of eight engineers at a company that sells billing software. Three weeks ago, while Kenji was on leave, Marta reworked part of his payments retry code to unblock her checkout feature and merged it after another teammate approved it. On his first morning back, Kenji reverted the change and posted in the team channel: “This is why we ask code owners first.” Since then the two of them talk only in pull request comments. Today you have a one-on-one with Kenji.
What you know
- Marta’s change passed review and all tests. Nobody asked Kenji.
- Pull requests between the two of them now wait two days or more for review.
- The payments release has slipped by a week.
- Two other engineers have asked not to be paired on the payments service.
Your goal
- Hear Kenji’s side without taking it.
- Separate what happened from what each of them assumes about the other.
- Agree on a plan: a dated meeting with both of them and clear rules for reviewing payments code.
The AI character
Meet Kenji. To him, this is about respect.
- His role
- Senior backend engineer, six years on the team. He built the payments service and still answers most of its on-call alerts.
- The other side
- Marta joined eight months ago as a full-stack engineer. Her checkout feature depends on his service, and her deadline is close.
- What he cares about
- A payments service that doesn’t break at night
- Being asked before his code changes
- Credit for what he built
- What he won’t say first
- He was passed over for tech lead last year and suspects Marta is next in line.
- He knows the public revert embarrassed her.
- The code Marta changed had a bug he had been meaning to fix.
- How he reacts
- You defend MartaHe shuts down: “So this is my fault.”
- Questions about the serviceHe relaxes and explains in detail.
- “You two just need to get along”Sarcasm, then he asks what that means in practice.
- Starting mood
Defensive and tired. Expects to be blamed before he is heard.
What is scored
Five criteria. Each one scored from what you say.
Once the one-on-one with Kenji ends, the AI checks every line you said against each criterion and quotes the ones that won or cost you points. Empathy counts through your words: whether you put his concern in your own words before you move on. Your company’s conflict guidelines can replace these criteria.
- 01
Stays neutral
What good looks likeAsks for his account first and says Marta will be heard too, without agreeing or disagreeing with either of them.
Common missExplains in the first minute that Marta was only trying to help.
- 02
Separates facts from assumptions
What good looks likePuts his concern in his words, then asks what Marta actually said or did, apart from what he believes she wants.
Common missArgues about her motives with him.
- 03
Names the impact on the team
What good looks likeDescribes what the standoff changes for everyone: reviews that wait, a release that slipped, teammates avoiding the code.
Common missMakes it about personalities and asks them to “be more professional.”
- 04
Gets him to own his part
What good looks likeAsks what he would do differently, so he names the public revert himself.
Common missLectures him about the revert, or lets him leave with all the blame on Marta.
- 05
Agrees on a plan
What good looks likeEnds with a dated meeting that includes Marta and a concrete rule to settle there: who reviews payments code and how fast.
Common missCloses with “Can you two sort it out between yourselves?”
How the conversation can go
From “ask her, not me” to one meeting with both of them.
Kenji · AI backend engineer
If this is about Marta, ask her why she merged my code while I was on leave.
You
I’ll talk to Marta too. Right now I want your side. What did you find when you got back?
AI coach · Stays neutralYou didn’t defend either of them. You asked for his account first.
Kenji · AI backend engineer
Half my retry logic rewritten. Nobody asked me. She wants the payments service, that’s obvious.
You
So the change went in and nobody checked with the person who built it. I can see why that stung. When you say she wants the service, what has she said or done that tells you that?
AI coach · Separates facts from assumptionsYou put his concern in your own words, then pulled what happened apart from what he assumes about her.
Kenji · AI backend engineer
Nothing directly. She just keeps changing things I built.
You
Here’s what I see from the outside: your reviews of each other wait two days, and the payments release moved a week. That’s why we’re talking. Looking back, what would you do differently?
Kenji · AI backend engineer
Fine. Reverting it in the team channel wasn’t my best moment.
AI coach · Gets him to own his partHe named his part himself. Don’t add to it; move to a plan.
You
Thanks for saying that. Can the three of us meet on Thursday and agree on who reviews payments code, and how fast?
Kenji · AI backend engineer
Thursday works. I’ll bring the review rules we used last year.
AI coach · Agrees on a planA dated meeting with both of them and one rule to settle there. You never had to decide who was right.
Conflict resolution · 5 criteria84/ 100
The plan is dated, but the cost to the team came late, after he had dug in. Name it earlier, then try the level where Marta joins.
Difficulty
Three versions of Kenji. Start one-on-one, finish with both in the room.
Level 210 min
Defensive
The default version. He blames Marta and mentions the tech lead decision only if you ask what this is really about.
- Blames the other side
- Hidden worry about promotion
Level 314 min
Both in the room
Marta joins as a second AI character. They interrupt each other, and both ask you to decide who was right.
- Two AI characters
- Interruptions
- Pressure to pick a side
Related scenarios
Practice the conversations around it.
Omar · AI senior claims handler
ManagersInsurance
Giving feedback to an employee
Your most experienced claims handler skipped a file note, and a customer escalated. Name what happened, find the cause and agree on one change.
Difficulty: IntermediateDuration: 8 min
Nia · AI account lead
ManagersB2B software
Performance review with an underperformer
Your account lead missed all three of her goals, two of them after an acquisition, and thinks the rating ignores why. Deliver it, hear her out and agree on a plan with dates.
Difficulty: IntermediateDuration: 10 min
A conflict between two people on your team is one of many hard talks a manager has to lead. In AI practice for the conversations managers put off, they also rehearse underperformance, a declined promotion and a restructuring.
Browse all sales, service and leadership scenariosFAQ
About this scenario.
How do you resolve a conflict between two team members?
Talk to each person separately first and hear their side without taking it. Separate what happened from what each one assumes about the other, name what the conflict costs the team, and ask each of them what they would do differently. Then bring both together to agree on a plan: who decides what, and how they will work from now on.
Can we run this conflict resolution role play without AI?
Yes. A colleague takes Kenji’s card and plays him tired and sure he’ll be blamed, someone else is the manager, and an observer keeps score. Neutrality is the hardest criterion to judge live, so the observer should write down every line that sounds like taking Marta’s side. For the hardest level, a fourth colleague plays Marta.
How does the AI score empathy and neutrality?
From the transcript. It checks whether you put Kenji’s concern in your own words, ask before you judge and avoid siding with either person, and quotes the exact lines each score rests on. Your tone and emotions are not analyzed. Only Kenji has a mood, as a character in the scenario.
Can we change the team, the conflict or the people?
Yes. Swap the engineers for your own roles, the reverted merge for a conflict your managers really face, and the review rule for your own code review process or team agreements. The scenario builder sets that up, and your leadership model can supply the criteria.
A quiet standoff won’t fix itself. Practice the conversation first.
In a short demo, play the manager opposite Kenji yourself, or bring the conflict your managers find hardest to raise.