AI in Coaching: What Actually Works (and What Doesn't)
Most AI tools in coaching are solving the wrong problem. Here's what the evidence says about where AI genuinely adds value in coach development.
Coaching is fundamentally a human relationship. The breakthroughs happen in the space between questions, in the silences, in the moments a coach chooses to stay with discomfort rather than move on. No AI changes that.
What AI can change is everything that happens around that relationship — the documentation, the pattern recognition across dozens of sessions, the ability to surface a growth edge that a coach has been missing for six months.
Where AI adds genuine value
Longitudinal pattern recognition. A coach working with twenty clients across twelve months generates thousands of data points. No human can hold all of that simultaneously. AI can surface patterns — recurring themes a client returns to, moments where a coach's questioning style shifts, competencies that are developing versus stalling — that would otherwise stay invisible.
Structured reflection prompts. Evidence from reflective practice research suggests that structure improves depth. AI-generated prompts tied to specific session moments (not generic journaling questions) can help coaches move from surface description to deeper analysis.
360 synthesis. Synthesising stakeholder feedback manually is slow and prone to the same biases it's meant to surface. AI can identify themes across multiple respondents without flattening individual voices.
Where AI falls short
Relationship context. An AI doesn't know that this client just went through a divorce, or that the rapport built over eight sessions makes a challenging question land differently today. Context lives in the coach's head, not the transcript.
Nuance and tone. AI reads words. It misses the pause, the laugh, the moment a client's energy changed before they said anything. Coaches catch these things. AI currently doesn't.
The "what matters" judgment. Even the best AI observation is a hypothesis. The coach who was in the room decides what's true. That judgment can't be automated — and shouldn't be.
The right model
AI as map, not driver. It helps you see terrain across sessions that would otherwise be invisible. The coach decides where to go. That framing — human-accepted, AI-surfaced — is the only one that holds up under scrutiny.