Questions Managers Should Ask AI Before a Difficult Employee Conversation

Preparing to lead, not asking for a script

Every manager eventually faces a conversation they would rather not have. An employee's performance has slipped. Someone's behavior is affecting the team. A previously reliable employee has become disengaged. Two people aren't getting along. An employee expects a promotion the manager doesn't believe they're ready for.

Today, managers have something previous generations didn't: an AI tool available at almost any moment to help them prepare. That can be extremely useful. A manager can ask AI to suggest questions, identify assumptions, consider alternative explanations, anticipate reactions, or help organize a conversation.

But there's an important difference between using AI to prepare yourself to lead and using AI to tell you what to say. The first can make you a better manager. The second can make a difficult conversation sound polished without necessarily making it effective.

Preparation can be assisted by AI. Presence cannot.

Why this matters now

Imagine a manager typing: “I have a difficult employee who is always negative. Tell me what to say to her.” AI will probably produce something. It may even sound very professional.

But look at everything contained in that prompt. The employee has already been labeled difficult. Her behavior has been characterized as always negative. The manager has asked for a script rather than help understanding the problem. AI is now working inside the manager's interpretation of the situation.

A better starting point might be: “I believe an employee's behavior is negatively affecting the team. Help me identify what I should understand and what assumptions I should examine before I speak with the employee.”

Now AI isn't being asked to confirm the manager's conclusion. It's being used to improve the manager's thinking. That's a much more valuable use of the technology.

Prompts that widen your view

Before a difficult conversation, use AI to help you see more of the situation — not less. Here are several questions managers can use.

“What assumptions might I be making?” This may be one of the most valuable questions a manager can ask. If someone is missing deadlines, perhaps they're disorganized. Perhaps. But expectations may be unclear, priorities may conflict, the workload may be unrealistic, the employee may lack training, or another department may be creating delays. The point isn't that the manager's interpretation is wrong — it's that it shouldn't automatically be treated as fact.

“What information should I gather before this conversation?” What expectations were communicated? What happened, and when? Is this an isolated incident or a pattern? Has the issue previously been discussed? What specific examples can you provide? What impact has the behavior or performance had? A difficult conversation becomes considerably more productive when the manager enters with specific observations rather than general conclusions.

“What questions could I ask instead of making assumptions?” Instead of “Why can't you ever get this done on time?” you might begin with “I've noticed the last three reports were submitted after the deadline. Can you help me understand what's getting in the way?” AI can help generate open-ended questions. The manager decides which ones fit the situation.

“What might this situation look like from the employee's perspective?” You don't have to agree with that perspective. You simply need to consider that another one exists. The employee may perceive conflicting expectations, believe priorities changed, not understand the impact of their behavior, or believe they're performing well. Or the manager's original assessment may be entirely correct. The purpose is to enter the conversation prepared to listen as well as speak.

“How could my own management approach be contributing to this problem?” Have expectations been clear? Has feedback been timely? Did the employee receive adequate resources? Has the manager avoided addressing the issue until frustration accumulated? Sometimes the answer will be that you've done everything reasonably expected of you. Sometimes it won't. Effective managers need to be willing to consider both possibilities.

“What outcome should I reasonably want from this conversation?” Greater clarity? A change in behavior? Improved performance? Agreement about expectations? A development plan? A commitment to specific actions? More information before determining the next step? Knowing the desired outcome changes the conversation.

“What reaction should I be prepared for?” The employee may agree, be surprised, become defensive, provide information the manager didn't know, become emotional, or challenge the manager's version of events. AI can help a manager think through possible responses and practice remaining calm and focused. But remember: AI can simulate a reaction. It cannot predict a person.

Prompts that narrow it dangerously

Some prompts do the opposite. Instead of broadening the manager's thinking, they encourage AI to validate a conclusion that has already been reached.

Be cautious with questions such as: “Why is this employee so difficult?” “How do I prove that this employee isn't performing?” “Tell me why this person isn't leadership material.” “How do I convince this employee that they're wrong?” “Should I fire this employee?”

The problem isn't simply that AI might provide a poor answer. The problem is how the question frames the situation. If you ask AI to justify your conclusion, it may generate arguments supporting that conclusion. That's confirmation, not leadership.

Instead of “Why isn't this employee ready for promotion?” ask “What evidence should I consider when evaluating whether someone is ready for greater leadership responsibility?” Instead of “How do I deal with a negative employee?” ask “What should I understand before addressing behavior that I believe is affecting team morale?” Instead of “Should I terminate this employee?” involve the appropriate HR and, when necessary, legal professionals. Some decisions shouldn't be outsourced to a prompt.

Protect employee confidentiality

There is another critical boundary. Don't assume that because information can be pasted into an AI tool, it should be.

Managers routinely handle sensitive information about performance, compensation, health, leave, disciplinary matters, complaints, investigations, and other employment issues. Follow your organization's AI-use, privacy, HR, and information-security policies.

Where AI use is permitted, frame situations without unnecessary identifying or confidential information. A manager generally doesn't need to tell an AI tool an employee's name to ask: “What questions could help me understand why a previously reliable employee has begun missing deadlines?” The management problem can often be explored without exposing the employee.

Know when to stop prompting and call HR

AI can help prepare managers for many routine leadership conversations. Other situations require professional guidance.

Managers should know their organization's escalation requirements and involve HR or other appropriate professionals when circumstances involve issues such as discrimination, harassment, retaliation, accommodations, protected leave, investigations, workplace safety, threats, significant disciplinary action, termination, or other potentially sensitive employment matters.

The manager's AI conversation should never become a substitute for organizational policy or qualified HR and legal advice.

Deciding what you will actually say

After using AI to prepare, close the tool. Then ask yourself: what do I actually need this employee to hear from me?

You probably don't need a perfectly worded five-paragraph script. You need a clear beginning. For many conversations, that means being able to explain what I observed, why it matters, what I need to understand, and what needs to happen next. Then listen.

That's the part AI can't do for you. It can't sit across from your employee. It can't notice that the person suddenly looks confused. It can't hear the hesitation before an answer. It can't realize that something you believed was a performance issue is actually a communication problem. It can't decide that this particular moment requires another question rather than the next sentence in the script. And it cannot build the relationship that allows difficult feedback to be heard.

Preparation can be assisted by AI. Presence cannot.

What to do next

If managers in your organization are using AI — and many probably are — don't make the first organizational response a list of prohibitions. Teach them how to use it better.

Give managers a simple framework: use AI to question your thinking; use AI to prepare, not decide; protect employee information; know when HR or legal needs to be involved; speak in your own voice; listen to the person in front of you.

And perhaps give managers one final prompt to use before an important employee conversation: “What might I be missing, and what should I be prepared to learn from this employee?” Instead of giving the manager an answer, it prepares the manager to enter the conversation with curiosity, clarity, and judgment.

Better prompts shouldn't replace better managers. They should help create them.