There’s a version of a “good leadership conversation” that’s easy to picture: the manager listens carefully, asks thoughtful questions, acknowledges the other person’s perspective, and works collaboratively toward a solution. The employee feels heard, the manager feels effective, and everyone leaves the conversation on relatively good terms.
Sometimes, that’s exactly what good leadership looks like.
But sometimes good leadership means telling someone clearly that their performance needs to change. It might mean refusing a request, communicating a decision that isn’t negotiable, establishing a boundary, or recognizing that an issue needs to be escalated to someone else.
And sometimes a well-led conversation ends with another person still frustrated, disappointed, or in disagreement.
Good leadership doesn’t always sound the same. And if we’re going to use AI to help people practice leadership, our simulations shouldn’t pretend it does.
The problem with a “good conversation” formula
Generative AI is remarkably good at conversation. It can listen, ask questions, express empathy, reflect someone’s language, and generate thoughtful responses in real time. Those capabilities make AI incredibly useful for creating realistic conversational practice.
But they can also create a trap.
If we aren’t thoughtful about how simulations are designed, AI-powered leadership practice can begin to reinforce a fairly predictable model of “good” communication: be warm, ask open-ended questions, validate feelings, avoid being too directive, find common ground, and work toward agreement.
None of those behaviors are inherently wrong. In many situations, they’re exactly what a leader should do. The problem comes when the same approach is rewarded regardless of the situation.
Eventually, learners figure out the pattern. They become better at producing the type of conversation the simulation seems to want.
But are they actually getting better at leadership?
Different moments require different leadership moves
Imagine an employee comes to their manager because they’re struggling with a problem. The manager has done this work before and knows exactly how to solve it. It would be faster (and probably easier) to simply give the employee the answer.
But if the goal is to develop the employee’s capability, that might not be the best move. Asking what they’ve already tried, what options they’re considering, or what they think the next step should be could create space for the employee to work through the problem themselves.
In that moment, coaching rather than solving may be exactly what good leadership requires.
Now change the situation.
An employee has repeatedly ignored an important team standard. The manager understands the context, the expectation has already been established, and the behavior continues.
Another round of open-ended questions may not be particularly useful. The manager may need to be clear about the standard, explain what needs to change, and establish what happens next.
Neither conversational style is inherently better. The situation determines what leadership requires.
The same is true across countless other management moments. A leader might need to ask, coach, investigate, clarify, decide, negotiate, push back, or escalate. Leadership requires a repertoire, not a single preferred style.
Sometimes the right answer is “no”
Consider a manager whose leader asks them to commit their team to an unrealistic deadline.
The manager could be agreeable. They could promise to “do their best” and avoid creating tension in the moment.
But that might simply push the problem downstream to their team.
A more responsible response could be to make the tradeoff explicit: We can prioritize this new request, but doing so means moving the other deliverable. Which should take priority?
That response may create more friction than simply saying yes, but reducing friction isn’t always the goal of a leadership conversation. Sometimes the goal is clarity, particularly when avoiding tension now will create a larger problem later.
This is one reason leadership practice becomes less useful when every scenario subtly pushes learners toward the same conversational destination.
A good conversation doesn't always end happily
The same issue appears when we evaluate the outcome of a conversation.
Imagine a manager has to communicate a difficult organizational decision. The decision has already been made, and the manager doesn’t have the authority to change it.
The employee disagrees. They explain why they think the decision is wrong. The manager listens, acknowledges the impact, answers what they can honestly answer, and makes clear which parts of the decision are (and aren’t) open for discussion.
The employee still leaves disappointed.
Was that a failed conversation?
Not necessarily.
Leadership isn’t always about getting another person to feel better or agree with the outcome. Sometimes it’s about creating clarity. Sometimes it’s maintaining a standard or communicating a difficult decision honestly. In other situations, a conversation might appropriately end in disagreement, escalation, or even a decision to pause until more information is available.
That’s an important consideration when designing leadership simulations. Emotional harmony can be one positive outcome, but it can’t be the universal definition of success. The Practice Ground editorial framework explicitly recognizes that credible simulations need room for different legitimate leadership moves and different emotional outcomes.
This is where simulation design gets harder
If the goal of AI roleplay is simply to create a believable conversation, much of this doesn’t matter. Give the AI a persona, establish a scenario, let the learner interact with it, and provide some feedback afterward.
But behavioral practice asks more of the experience.
A simulation has to reflect the fact that different situations require different capabilities. The behaviors that make sense when coaching an employee may not be the same behaviors needed to hold a high performer accountable. Leading someone through uncertainty is different from communicating a final decision. Managing conflict between employees is different from pushing back on an unrealistic request from above.
That means learners shouldn’t be able to discover one “winning” conversational style and apply it successfully everywhere.
The objective isn’t to teach managers to reproduce the perfect sentence. It’s to give them opportunities to make decisions in context.
What does this particular moment require from me?
That might mean asking another question. It might mean coaching instead of solving. It could mean being more direct, establishing a boundary, making a decision, or recognizing that the issue belongs with someone else.
Those are judgment calls, and judgment is difficult to build from a slide.
Practice should build a repertoire, not a script
Frameworks, examples, workshops, and coaching all play valuable roles in leadership development. They can help managers understand what good leadership looks like and give them language for situations they’re likely to encounter.
Practice adds something different: the opportunity to use that knowledge when another person is responding in real time.
A learner makes a choice. The conversation changes. New information emerges. The other person pushes back. An approach that seemed reasonable doesn’t land the way the learner expected.
Now they have to decide what to do next.
That uncertainty is part of the value. Real conversations don’t unfold according to a script, so leadership practice shouldn’t depend on one either.
Over time, varied practice can help managers develop a broader repertoire. A first-time manager might practice coaching, delegation, accountability, conflict, managing up, and communicating difficult decisions. A leader navigating AI change might encounter uncertainty, experimentation, governance, evidence, changing roles, and competing priorities.
Different situations create opportunities to practice different leadership moves.
Good leadership practice should reflect real leadership
The goal of leadership development shouldn’t be to produce managers who all sound the same.
It should be to help people become better at recognizing what a situation requires and choosing an appropriate response, even when there isn’t one perfect answer.
That’s the philosophy behind Practice Ground by Virbela.
Through AI-powered simulations, people can rehearse consequential workplace conversations, experience how those conversations respond to their choices, receive feedback, and practice again before the stakes are real.
Because real leadership doesn’t come with a script.
Good leadership doesn’t always sound the same. Good leadership practice shouldn’t either.
