Skip to content

AI for managers: conversations, feedback, rules for the team

By the KI können editorial team7 min read

Written with AI and checked against the sources by a second AI. As of November 17, 2026. More in the notes on AI.

Leading people is largely talking and writing: preparing conversations, giving feedback, handing over tasks, explaining decisions. An AI can help you find the right words. What you think and decide about your people stays your business. And a team that uses AI needs clear rules from you.

In a meeting room, a team lead talks with an employee, her notebook and two cups on the table, a window and a plant behind them.

Key points

  • For conversations, the AI collects questions and a structure. Names and performance data stay out.
  • It phrases tough feedback concretely and respectfully if you say what you observed.
  • Don’t let an AI evaluate, assess or monitor anyone. Emotion recognition in the workplace is prohibited, except for medical or safety reasons.
  • Before the team uses an AI tool, the company settles the rules, with the works council where there is one.

What managers use AI for and what not

Management tasks and the role of the AI
TaskThe AI helps withStays with you
Employee reviewStructure, open questions, wordingImpression, assessment, promises
Critical feedbackConcrete, respectful wordingWhat you observed
Handing over a taskA clear description with goal and limitsWho gets it and why
Announcing a changeAn understandable message, likely questionsThe decision and its reasons
Assessing performanceNothingEverything

In the trial lesson you put your first instruction together and get feedback. No account, in a few minutes. Start the trial lesson

Preparing conversations without personal data

Lindsay leads a customer service team at an energy supplier. Once a year she has a development conversation with each of her twelve people. Preparing for them costs her an evening every time. Now she describes the situation to the AI without names and without figures from the performance report and has it suggest a structure.

Sample instruction
I’m preparing an annual development conversation. I’m a team lead in customer service.
Situation, without names: the person has been on the team for three years, is professionally confident, wants more responsibility, but often feels pressure with complaints.
Suggest a structure for 45 minutes and eight open questions that give them room to talk.
No questions about health, family or private life.
No assessment of the person.

Lindsay adapts the suggestion. She drops two questions because they sound too much like a textbook. What she thinks about the person’s performance goes into her own notes, not into the chat. Conversation notes belong in the company’s personnel records.

Phrasing criticism concretely and respectfully

Derek is a foreman in production at a plastics processor. A machine operator has filled in the handover to the next shift incompletely for the third time. Derek’s first sentence in his head: “This can’t go on.” True, but not helpful. He has the feedback rephrased.

Sample instruction
Help me phrase feedback to an employee that I’ll say in a conversation. Direct, without blame.
Observation: the handover log was incomplete three times in two weeks; the last time, the tool change was missing.
Consequence: the next shift spent an hour looking for the fault.
What I want: every handover complete; under time pressure, come to me briefly instead.
Write three or four sentences I can say, and one open question that lets him have his say.

The instruction has three parts: observation, consequence, request. The AI doesn’t need more than that, and it shouldn’t know more. The text is a suggestion for Derek’s own words, not a script to read out.

Handing over tasks so that no follow-up questions are needed

Jared leads a small IT team. When he hands out tasks, the same questions keep coming: by when, how thorough, who has a say. His notes are brief. The AI turns them into a description that he puts in the ticket.

Sample instruction
Turn my notes into a task description for a colleague. Sections: Goal, What done means, By when, What she decides herself, When she asks me.
Where my notes leave something open, write [to clarify] and ask me three questions about it at the end.
Notes: [notes]

The AI’s questions show Jared what he used to assume without saying. He answers them before the task goes out.

Announcing changes

Whitney runs a home care service. From next quarter, route planning will change, and she knows that will cause unrest. She has a message to the team drafted, along with a list of questions that are likely to come.

Sample instruction
Write a message to my team at a home care service. Honest, without sugarcoating, at most 200 words.
What changes: [change]. From when: [date]. Why: [reason].
What stays the same: [points]. When we’ll talk about it: [meeting].
Then, separately: list ten questions or worries the team will probably have, so I can prepare answers.

For Whitney, the list counts for more than the message: she prepares an honest answer to every question. Where she has none, she says so in the team meeting instead of promising something.

Rules for AI in the team

If your team already uses AI, maybe with private accounts, a ban in the hallway won’t help, but a rule will. The EU AI Act requires companies that use AI to take measures that support their people’s AI literacy. No particular certificate is required. As a manager, you make sure your team knows the rule.

  • Which AI is approved, with which account?
  • What may go in, and what never: customer data, personnel data, trade secrets?
  • Who checks results before they go outside?
  • Who do you ask when in doubt?

If there is a works council (Betriebsrat, the elected employee representation in German companies), it must be informed in good time about planned AI in work processes. It can bring in an expert for this, by arrangement with the employer. If a tool can monitor employees’ behavior or performance, the works council has a say. So don’t introduce a new tool in your team on your own, but talk to your manager, IT and, where applicable, the works council first. In schools, public authorities and church institutions, separate rules for staff councils or employee representation apply instead of the Works Constitution Act.

When the team hands in AI results

Sooner or later, someone on the team hands in a report, a quote or a concept that was clearly made with AI. That’s no reason for mistrust, as long as the rules are followed. What matters is whether the person checked the result and stands behind it.

  • Don’t ask “Did you use AI?” but “How did you check this?”
  • Expect the source for figures, facts and quotes, no matter who wrote the text.
  • Say openly where you use AI yourself. That makes it easier for the team to say so too.
  • If a mistake shows up, discuss the checking step, not the tool.

A clear line: evaluating and monitoring people

An AI doesn’t know your people. It only sees what you give it and fills gaps with assumptions. That’s why assessments, rankings and decisions on promotion, warnings or dismissal don’t belong in an AI chat. The decision stays with you, and you must be able to justify it.

  • Prohibited since February 2025 are AI systems that infer emotions from faces or voices in the workplace, except for medical or safety reasons.
  • High-risk are AI systems that help decide on hiring, promotion, task allocation or the performance of employees. Strict obligations apply to them from December 2, 2027, including for the company that uses them.
  • Inform beforehand: also from December 2027, the company must inform employees and their representatives before it uses such a system in the workplace. German law already requires informing the works council today.

More on risk levels and deadlines is in the guide EU AI Act for employees, more on references and job ads in the guide AI in human resources.

Before you say or send anything

  1. No name, no performance figure and no health information in the instruction.
  2. You made every observation in the text yourself.
  3. The text sounds like you. Otherwise rewrite it until it does.
  4. No promise you can’t keep.
  5. All [to clarify] replaced, all of the AI’s questions answered.

Model what you expect from the team

A team adopts the way its leaders work. If you write clear instructions yourself and visibly check results, you don’t have to preach checking. The learning app “KI können” practices exactly that, step by step and in the AI your company has approved. Try the trial lesson yourself before you recommend it to the team; you don’t need an account for it.

Sources

Information as of November 17, 2026. Laws and offers can change; this article is not legal advice.