Most of the conversation about AI in leadership development is happening in the abstract. People are predicting what AI might do, what it could replace, or how it will transform the field.
Over the past year, my team has been doing something different. We have been building AI leadership-development tools and testing them with leaders. The experience taught us something important about where AI creates real value, where it does not, and what separates truly developmental tools from gimmicks.
We built two.
The first is a pattern-recognition tool that analyzes how a leader diagnoses a situation and names their recurring blind spots. It sees when someone treats a symptom as a cause, leans on authority where influence is needed, keeps avoiding the difficult conversation, or misses the chance to develop the people around them. Most leaders experience these as isolated incidents. The tool shows them the pattern running underneath.
The second is an AI-powered leadership simulator. Leaders step intorealistic adaptive challenges - post-merger tension, competing co-leads, eroding trust, stakeholder conflict, resistance to change. They make decisions, watch the consequences unfold, and get feedback. As their capability grows, the scenarios get harder and the feedback gets sharper. There are no fixed scripts. The simulation responds to how the leader actually leads.
Building these changed my view of where AI belongs in this work.
Where AI Creates Real Value
AI earns its place in the parts of leadership development that have always been weakest: repetition, exposure, and feedback at the moment a decision gets made.
Most leaders never practice the hard situations before the consequences are real. They get one chance to lead a major reorganization. One chance to respond to a crisis. One chance to handle a breakdown of trust with a key stakeholder. The learning arrives after the fact, once the cost has already been paid.
A simulation changes the reality of practice. A leader can run the same hard scenario eight times and try different approaches - watch a move to centralize authority erode trust, see how avoiding conflict creates a worse problem later, feel the difference between moving too fast and moving too slow. Case studies teach concepts. Simulations build experience. The leader starts to recognize their own patterns before they meet the situation for real.
The pattern-recognition tool solves a different problem. Most leaders cannot see the consistency in their own behavior. Every situation feels new, every challenge looks like it has its own explanation. Growth usually depends on seeing the opposite: the leader who keeps solving the technical problem and ducking the adaptive one, who over-functions for the team, who believes they are empowering people while quietly holding on to control. Those patterns stay invisible because they show up across dozens of moments, not in any single one. A skilled coach can surface them. An AI tool developed by a skilled coach can also do it - and offer it at scale.
That value is real, and it is specific.
What Surprised Us
The biggest surprise had nothing to do with the technology. It came down to expertise.
As an experiment, we rebuilt the simulator on public models - Perplexity, Gemini, Claude, the free ChatGPT - using the same prompts. On the surface it looked fine. The models generated plausible scenarios and coherent feedback, and the conversations often sounded sophisticated. Then we pushed harder and the gap opened up. Every model could generate language. What none of them could do was reliably develop judgment: name the underlying developmental challenge instead of the surface behavior, tell a technical problem from an adaptive one, give feedback that built capacity instead of describing performance.
The difference was the intellectual property sitting behind the model. Our tools run on our leadership frameworks, our diagnostic methods, our authority-calibration models, our developmental-feedback approaches, and years of watching how leaders actually learn. The AI extends that work into a new medium. Strip the expertise out and you have a tool that sounds developmental and produces almost no development. That distinction gets very little attention in the current conversation about AI, and it is the whole game.
What This Means
After a year of building, my conclusion is both more optimistic and more cautious than when we started. AI can widen access to practice, speed up feedback, and surface patterns leaders cannot see on their own. Those are real gains. But the value lives in the expertise we build into the tool - the model only extends it. A tool trained on shallow expertise produces shallow development faster. A tool built on decades of insight extends that insight into something truly new.
Explore more leadership resources from KS Insight →