Mentoring and coaching are very different, and AI affects both.

Warren Buffett, the American investor and philanthropist who built one of the world's most successful companies by investing in good people and then getting out of their way, has a simple view of leadership development.

Buffett says, "find the .400 hitters and then don't tell them how to swing."

For those unfamiliar with baseball, the analogy may not be obvious. In Full House: The Spread of Excellence from Plato to Darwin, S.J. Gould examines the extreme rarity of the .400 hitter. That's one of the most elusive feats in Major League Baseball. The last player to hit .400 was Ted Williams of the Boston Red Sox in 1941 (he actually hit .406). Eighty-five years later, the "Splendid Splinter" is still the only one to have done it.

Buffett's view is that you identify the very best people and get out of their way. Bill Gates expresses a very similar view in his introduction to Tim Gallwey's The Inner Game of Tennis: The Classic Guide to the Mental Side of Peak Performance. He argues that the single biggest impediment to success is often you. That may well be the case for the organization, too. Gallwey was a key figure in the coaching movement and understood that performance is often constrained more by how we think than by any lack of knowledge or ability.

Fear of failure, self-doubt, and overreliance on others' direction can all significantly impede performance. You can certainly address these, but the sheer amount of time and money organizations spend on leadership development suggests it's not easy.

This brings me back to the distinction we noted earlier: mentoring is not the same as coaching. Mentors pass on knowledge, while coaches help the individual discover it.

This is a very important distinction. Mentoring draws on the experience and expertise of those who have been there before. Coaching is about asking the right questions to challenge assumptions, provoke thought, and enable the individual to arrive at their own solutions. Both are, of course, very valuable.

The good leader knows when to mentor and when to coach.

This isn't just a theoretical point. Research is clear about the value of coaching in developing educational leaders. A 2025 systematic review in Frontiers in Education examined coaching as a vehicle for school leaders' professional learning. A 2026 Learning Policy Institute report (The Principal Effect) makes a similar point and recommends investing in coaching, mentoring, professional networks, and opportunities for self-directed learning as key elements of an effective approach to principal development.

However, there's a major issue with that. Developing leadership capacity across school districts has always been highly inconsistent. It's largely a matter of where you are in the structure and, more importantly, who you report to. The principal is expected to develop the assistant principal, the director to develop the coordinator, and the assistant superintendent to develop the principal and the director (at least that's the theory).

Throughout my progression to the position of superintendent, I was extremely fortunate to have excellent supervisors. A number of them were very good mentors and advisers, but only one actually provided coaching. His name was Dan.

Dan was my principal when I was a middle school teacher. He promoted me to interim assistant principal shortly after I earned my PASC. He had a background in athletic coaching and was a big fan of Gallwey and the GROW coaching model. More than that, he understood the fundamental principle that so many administrators still don't grasp, namely that developing leaders is not the same as managing them.

Dan was never judgmental. He was far more interested in understanding the dynamics of the situation than in simply judging the outcome. If I came to him with a problem, he would rarely give me a solution. He'd ask me about my instincts and pose questions that would make me look at my own instincts, tendencies, and the consequences of the decisions I was going to make.

As a novice admin, I didn't always trust my instincts and wasn't sure whether they were reliable. Dan made it clear that the key to leadership development is determining whether your instincts will be a good thing or a bad thing. He didn't think for me at all; he just helped me think better myself.

This was exemplified coaching. Now consider having access to that experience at any time. An entity that never gets tired of questions, never needs to be rescheduled, and is always patient with the same questions being asked over and over again.

This is where AI comes in.

AI has moved far beyond answering questions, summarizing documents, and drafting emails. People are now using it for advice, reflection, personal development, and even emotional support. In a June 2026 HBR article, "How People Are Really Using AI in 2026," Marc Zao-Sanders reports that therapy and companionship were the two most frequently cited use cases in his analysis of online chatter about generative AI for the second year in a row.

This is quite significant, really. In a tech revolution that was initially heralded for its ability to perform the kinds of complex cognitive tasks we associate with human intelligence, one of the most common applications today is simply having a chat.

People don't just want AI to think for them; they want AI to think with them. It's crucial to emphasize that using AI as an emotional support tool is not the same as receiving therapy from a properly qualified professional, and that interacting with an AI chatbot is not, in itself, coaching. The APA has warned against using general-purpose AI as a substitute for proper mental health care, while recognizing that well-designed tools could, in principle, serve as an adjunct.

All of that is perfectly understandable given the appeal of AI. It's on demand, it's fairly cheap, and it provides a conversational space where people can go through the motions of a difficult conversation, consider alternative viewpoints, and reflect on their own experience, all without the immediate threat of being judged by someone else.

A newly appointed principal about to have a difficult conversation with an angry parent, an assistant superintendent dealing with an underperforming admin, or a superintendent navigating a broken board relationship could use AI as a mentor by asking for examples, research, strategies, and information. They could also use it as a coach to challenge assumptions, ask the right questions, explore alternative explanations, and work through the best course of action.

This underscores the key difference. When AI provides answers, it acts as a kind of mentor. When it facilitates self-inquiry, it can do some of what a coach would do.

However, the question remains: can AI truly provide coaching?

Research offers a qualified yes, with significant caveats. A 2026 RCT by Erik de Haan, Nicky Terblanche, and Kenneth Nowack compared human coaching to AI chatbot coaching and found meaningful gains in the human coaching group but not in the AI group. A second 2026 study examining manager responses to the same coaching approaches (human and AI) identified what the authors describe as the 'reflection penalty': participants were much more receptive to the human coach asking reflective questions and much more likely to want the AI to give them an answer.

These results suggest that people prefer a human coach to provoke their thinking, while often expecting a machine to give them the answer. This observation warrants careful consideration.

One of the greatest risks of using AI as a coach is that it can become too good at helping you avoid the very cognitive work that coaching is supposed to be about. If you ask AI what to do, it will generally do that. If you ask it to agree with your assessment of a difficult colleague, it will probably be very accommodating unless you specifically tell it not to be. Neither response builds leadership capability.

In other words, AI can become an extremely sophisticated enabler of our own worst instincts.

But emerging research points in a far more promising direction. In 2026, Terblanche and Rajashi Ghosh proposed a hybrid intelligence model for leadership coaching that brings together the cognitive power of AI with the relational intelligence of the human coach. Rather than viewing them as alternatives, the model recognizes that both can contribute to real leadership development.

This is where I see the real opportunity, especially for educational leaders. As I continue to build my coaching practice (I am a ICF-certified executive coach) I am finding great value in emerging AI technologies, especially when I am working with aspiring and sitting admin to help prepare them for their next position.

Imagine a school district that pairs up an emerging and an incumbent administrator with an experienced executive coach and an AI coaching partner. The human coach sets goals, builds the relationship and trust, challenges the person's assumptions and beliefs, and provides the context that comes from having led schools and districts. Between coaching sessions, the admin uses the AI to work through a problem, rehearse a conversation, test a decision, or reflect on the consequences of an action.

With the administrator's permission and appropriate confidentiality safeguards in place, those reflections can inform the next conversation with the human coach. The coach can then help the administrator identify blind spots, challenge AI-generated assumptions, and distinguish what sounds sensible from what is actually sound in the circumstances.

Think of an assistant principal in the middle of a tricky personnel issue. The AI mentor can explain the relevant leadership principles, help you generate questions for a conference, and help you collate the facts. The AI coach can ask you why you're not prepared to have the conversation with the employee, what assumptions are at play in your mind, and what other interpretations of the situation are possible. The human coach can then draw on their own experience, relational knowledge, and understanding of the district's culture to help the administrator make sense of the situation and their own response. The human coach doesn't need to be available every time the admin runs into a problem. The AI partner doesn't need to pretend to have human judgment or experience. The admin doesn't have to choose between them.

The outcome could be better than either approach on its own: a continuous developmental relationship that brings the accessibility of AI together with the judgment, empathy, and accountability of a highly skilled human coach.

I'd say that's particularly true in public education, where leadership decisions are rarely made in a vacuum. Context is everything. Relationships are everything. Organizational history is everything. And at times, the technically correct answer is, in fact, the wrong answer for a particular school community.

An AI system might help the principal understand the problem. A good coach can see why the principal is having such difficulty facing it. Together, they can help the principal build both the knowledge and the judgment to move forward.

That doesn't mean we should run a human-AI partnership without guardrails. Confidential personnel information, student records, and similar data require protection. AI can produce incorrect information, amplify bias, and offer advice with misplaced confidence. Coaching must be clearly separate from performance review if we expect administrators to reflect honestly on their own shortcomings.

Access to an AI chatbot is not, in itself, a coaching program. You need thoughtful integration, well-defined goals, clear expectations, human involvement, and demonstrable evidence of changed leadership practice. The bar isn't about how many times an admin talks to a coach (human or otherwise). The bar is whether the admin becomes a more skilled, more reflective, and more effective leader.

That brings me back to Dan. I wonder what he would have made of AI. I think he would have been quite interested in the prospect of it helping the aspiring admin work through a problem in his head before actually bringing it to him. I suspect he would have been quite reluctant to let it do all the work, though. He knew that the process of working through a problem was often more important than getting to the right answer immediately.

Dan didn't make me reliant on his judgment at all. He helped me to form my own. And there you go, isn't that the point?

The future of executive coaching in education shouldn't be a battle between human and artificial intelligence. It should focus on how the two can work together to develop better leaders. AI can make coaching more accessible, more continuous, and potentially more affordable. Human coaches can still provide the relationship, professional judgment, and accountability that AI cannot currently deliver with any degree of reliability.

I would love to hear how you are using AI as a tool for personal and professional development. I find it quite useful for both.