Mediating between two reports

When two reports clash, the quickest fix is to separate them: different shifts, different tickets, less contact. Sometimes that's right, but usually it hides the conflict rather than settling it.

Mediation is the alternative: the lead hears each person alone, then brings both together around the shared problem, and they find the next step themselves. The lead runs the process; the two people own the solution, which makes it far more likely to hold.

At Klasna, Bohdan wrote "A machine wrote this." under one of Dmytro's replies, and Dmytro asked for shifts without Bohdan. Separating them looks easy, since Dmytro's contract ends in two weeks. But the quarrel is about how the team uses the AI draft assistant and who checks its drafts, and that question stays after Dmytro leaves. Olha adds context: last month an AI draft promised a family a refund outside the policy, and Bohdan had to call them back. There's a real problem under the comment.

The core tool of mediation is separating positions from interests. A position is what someone says they want: "put me on shifts without Bohdan", "seasonal staff shouldn't use the AI assistant at all". An interest is the reason underneath: a need, a worry, a goal.

Positions are usually opposite; interests often aren't. Dmytro's position (separate shifts) protects interests like "I don't want to be corrected in front of the whole team" and "I need to keep up with the queue while I'm still learning the policies". Bohdan's comment protects "families mustn't get promises we can't keep". None of those interests contradict each other. Both men care about families getting correct answers quickly.

So in every conversation, listen through the demand to what it protects. A useful question: "What would that give you?" or "What worries you if we don't?" Once the interests are on the table, two opposite positions can often be reconciled with an option nobody had proposed.

Mediation starts with separate talks: about twenty minutes on a call with each person. Ask what happened, what matters to them, and what they'd need. Listen, summarize back what you heard, and don't judge or promise anything yet. Each person gets heard without an audience, and you collect the interests you'll build on.

At Klasna, Dmytro says: "Without the drafts I'm twice as slow. And his comments are public." Bohdan says: "I'm not his teacher, but I won't let a family be lied to." Underneath, both want the same thing: families getting correct answers without long waits. That shared interest becomes the problem you put on the table together.

Watch the power imbalance: ten years in the field against three months. In the joint talk it can make one voice count more. Agree the ground rules up front, give each person the same time, and let the quieter or less senior person speak first, so their view isn't shaped by the other's.

A mediation between two reports follows six steps, and the order matters.

1. Talk to each person separately, so each is heard without an audience.
2. Agree the ground rules for the joint talk: equal time, no interrupting, talk about the work, not the person.
3. Name the shared problem, not a culprit: "families need correct answers without long waits", not "Bohdan's comment".
4. Each says what matters to them, and the other restates it in their own words. Restating proves the interest was heard and often softens the room.
5. Collect options before choosing, so the first idea doesn't win by default.
6. Write down the agreement with a review date, so the talk turns into something you can check.

Skipping steps is the usual failure. Jumping to options before both feel heard gets polite agreement that falls apart in a week. Ending without a written agreement leaves each person remembering a different deal.

In the joint talk, the first idea that sounds workable tends to win, simply because it came first and everyone wants the awkward conversation over. Resist that. Ask for at least three options before anyone weighs them.

For Bohdan and Dmytro the options might be: AI drafts allowed for scheduling and package questions, refund replies written by hand; Bohdan reviews Dmytro's refund replies privately for two weeks; a short checklist before sending any draft; comments by direct message, not in shared tickets. Several can combine.

Then choose against criteria both accept, drawn from the interests you heard: correct for families, fast enough for the queue, nobody corrected in public. Criteria turn "my idea against yours" into "which option meets what we both need". Know the limits of the room, too: Iryna wants anything that changes the refund policy or the assistant's rules for the whole team to come to her. The two can agree how they work together; a team-wide rule is a proposal for Iryna.

Halfway through a joint talk someone often says something sweeping: "That assistant is garbage, full stop." It's a position in its strongest form, and how the lead responds decides whether the talk stays on the problem.

The useful move is a question: "What have you seen go wrong?" It treats the statement as information, not an attack. Bohdan's answer, an AI draft promising a refund outside the policy, turns into an interest the whole room can work with: refund promises must be right. Add it to the list of what matters, next to Dmytro's need to keep up with the queue.

Three tempting responses make it worse. Agreeing to ban the tool to calm things down hands Bohdan a win and drops Dmytro's interest. Telling him the remark is unprofessional scolds him in front of the other side and makes him defend it. Letting Dmytro answer directly turns the call into a contest. The lead's job in the room is the process, not the verdict.

Mediation fits two people with a work disagreement who can both speak freely. Several situations look like a conflict between two reports and aren't that.

Harassment, threats, discrimination, safety or legal issues: don't mediate. A coordinator mocking a colleague's accent in private messages is harassment with a discrimination angle, not a disagreement. Record the facts, protect the person who raised it, and take it to Iryna the same day.

Fear: if one person is afraid to be alone on a call with the other, they can't speak freely. Keep the talks separate and escalate.

A health crisis: if someone shows signs of one, point them to professional help; that comes before any mediation.

A conflict you're part of: a disagreement between you and Olha about how the rotation works can't be mediated by you. Ask Iryna or a neutral colleague.

And when the clash comes from the setup, mediate and fix the rule. Here nobody owns the team's rules for the AI assistant, which is why the quarrel happened.

Remote mediation works, with a few rules. Hold the joint talk on a private video call, cameras optional, never in a team thread: a public channel gives the conflict an audience and makes backing down harder for both sides. Transparency means a clear record of what was agreed, not a public argument.

Afterwards, send both a short written summary: what you agreed, who does what, when you'll check. Each confirms in writing, and you follow up privately a week later to see whether it's holding.

And be honest from the start about what happens if they don't agree: "If you two don't agree by Friday, I'll decide, and here's how." That isn't a threat. It's clarity about who decides, which makes the joint talk more serious, not less.

How it fits together: mediate instead of separating, listen for interests, hear each side alone, collect options before choosing, know when not to mediate, and end with a written agreement. Next, Overload and burnout: when the pressure is on everyone at once.