Why AI Struggles with Leadership Coaching

In 2016, Google introduced AlphaGo Zero, a groundbreaking AI that made history by defeating the best human players in Go, an ancient and highly complex board game. AlphaGo's journey to mastery offers a fascinating look at how AI excels in closed environments with clear objectives, yet struggles when applied to more ambiguous, human-centered tasks—such as leadership coaching.

AlphaGo’s success was due to two key stages of learning:

1. Learning by Imitation: AlphaGo studied thousands of games played by expert human players to understand foundational strategies.

2. Self-Improvement: It played against itself repeatedly, receiving rewards each time it won, allowing it to continuously refine its strategies without human input.

In just 40 days, AlphaGo became the best Go player in the world. This rapid improvement was possible because Go is a closed environment with a straightforward “reward function”—you either win or lose. This structure allowed AlphaGo to play millions of games in real time, quickly learning which strategies worked best.

Why Large Language Models Like ChatGPT Struggle with Leadership Coaching

When we look at Large Language Models (LLMs) like ChatGPT in the context of leadership coaching, things get more complex. Initially, these models learn in a similar way to AlphaGo’s first stage, by imitating expert coaches through training on thousands of coaching conversations. However, leadership coaching isn’t as straightforward as Go or chess; there’s no clear “win” condition. Coaching outcomes are subjective, varying from one individual to another, and can’t be easily categorized as success or failure.

This lack of a clear reward function presents a significant challenge. Unlike AlphaGo, which could perfect its game strategies by self-playing millions of times, an AI coach cannot simply “practice” leadership coaching in the same way. In leadership coaching, the “game” isn’t clearly defined, and the feedback isn’t immediate or objective. There’s no straightforward scoring system to assess whether the AI’s responses are “right” or “wrong.” The nuances of tone, empathy, and situational context are difficult for AI to self-learn without human-like judgment and experience.

Understanding the Technical Limitations of AI in Coaching

Without a defined reward mechanism, AI faces several barriers when applied to leadership coaching:

• Subjectivity of Success: Coaching effectiveness varies by individual and situation. A strategy that works well for one person may not work at all for another. AI, with its reliance on patterns and probabilities, lacks the nuanced judgment to adapt its responses based on subtle emotional or situational cues.

  
  

• Lack of Immediate Feedback: In Go, each game ends with a clear result. But in coaching, “results” unfold over time and can’t be instantly evaluated. The subjective and evolving nature of coaching outcomes limits the AI’s ability to quickly learn from its “mistakes.”

  
  

• Inability to Self-Improve in Real Time: AlphaGo refined itself through self-play, but there’s no equivalent for coaching conversations. An AI like ChatGPT can’t practice coaching without clear, real-time feedback.

Unlike a board game, coaching isn’t repeatable in a closed loop; each interaction is unique and subjective, making it challenging for AI to evaluate and adapt.

If you or your team are hesitant to adopt AI for leadership coaching, the reason could lie in these technical limitations. Unlike Go, where a clear win condition allows for rapid self-improvement, leadership coaching is grounded in subjective data, lacking the structure that makes self-teaching possible for AI.

Why the Human Touch Still Matters in Coaching

While AI can be a valuable tool in many business areas, leadership coaching still requires a level of nuance, empathy, and contextual understanding that AI hasn’t yet mastered. Effective coaching relies on deep insights, emotional intelligence, and the ability to adapt to unpredictable human behavior—qualities that AI has yet to fully replicate. Senior leaders who understand these technical challenges can set realistic expectations, appreciating the value of human experience in areas that require complex interpersonal skills.

The Road Ahead

AI has made incredible strides, and advancements continue to bring new possibilities. However, in areas like leadership coaching, where success isn’t defined by simple metrics, the human touch remains irreplaceable. As AI technology progresses, we may see improvements in this area, but for now, leadership coaching is a domain that benefits from the unique capabilities of human intelligence, understanding, and adaptability.