48% of coaching clients are already using ChatGPT between sessions. Not because they want a chatbot. Because their coach isn't there, and something is. The question isn't whether AI enters coaching. It's whether you're in the loop when it does.
Coach-supervised AI is the answer to that question. It's a model where AI proposes coaching support - session notes, reflection prompts, progress check-ins - and the human coach reviews, approves, and controls every output before it reaches a client. Nothing is automated. Nothing goes out without the coach's action. The AI extends the coach. The coach remains in charge.
We spoke to 52 coaches across 11 countries before building anything. 91% said the same thing: if AI touches my client, I need to approve it first. Coach-supervised AI is built on that principle.
How is coach-supervised AI different from AI coaching?
These terms get used interchangeably. They shouldn't. They describe fundamentally different bets on the future of coaching.
AI coaching replaces the human coach. BetterUp's AI coach costs nothing per session. It also has zero context on your client's actual life, their relationship with their CFO, or the fact that "delegation" is really about a fear of irrelevance they mentioned in passing three months ago. That's not coaching. That's a chatbot with a wellness badge.
AI-assisted coaching is a marketing term that covers everything from using ChatGPT for session prep to enterprise platforms with vague "AI features." It tells you nothing about who controls the output or whether a human ever sees it.
Coach-supervised AI is specific. The AI proposes. The coach reviews. Nothing reaches the client without the coach's explicit approval. The coach sets what the AI can and cannot do. The coach can override, edit, or reject any output at any point.
| Approach | Who controls the output? | Coach in the loop? | Client relationship |
|---|---|---|---|
| AI coaching | The AI | No | Client ↔ AI |
| AI-assisted coaching | Varies | Sometimes | Unclear |
| Coach-supervised AI | The coach | Always | Client ↔ Coach (AI supports) |
| Chatbot coaching | The AI | No | Client ↔ chatbot |
The distinction isn't semantic. It determines who holds the coaching relationship. In coach-supervised AI, the answer is always the coach.
Why does supervision matter?
57.7% of coaches say AI chatbots can't deliver proper coaching (EMCC/Henley Business School, 2024). They're right - but not for the reason most think.
The issue isn't that AI is bad at language. It's excellent at language. The issue is that coaching isn't a language problem.
Coaching is relational. A session about delegation might actually be about a client's fear of losing control, which connects to their leadership identity, which connects to feedback they received six months ago. Your client doesn't tell the AI this. They tell you - sometimes without words. An AI that has processed session transcripts has data. It does not have the relationship.
Coaching is contextual. It happens within systems - organisational politics, team dynamics, career trajectories, personal circumstances. The coach holds the full picture. The AI sees one transcript at a time.
Coaching is ethical. The ICF and EMCC have clear standards around confidentiality, consent, and the primacy of the client's welfare. When AI generates coaching content without human review, these standards are at risk. When a coach reviews every output, the ethical framework stays intact.
"I need to know that nothing reaches my client that I haven't seen and approved. That's not a nice-to-have. That's the foundation of the relationship." - Leadership coach, London (from our 52-coach research)
Without supervision, you have automation. With it, you have augmentation. That's the entire difference.
Want to see coach-supervised AI in practice?
How does it work in practice?
Five steps. The coach is in the loop at every one.
- Session recording. The coaching session is recorded (with client consent) and transcribed. The transcript stays on EU-hosted infrastructure. Not US servers. Not used for model training.
- AI generates insights. Within minutes: session notes, emerging themes, developmental patterns across sessions, and reflection prompts tailored to what was discussed. Not meeting minutes. Coaching intelligence.
- Coach reviews everything. Every output, every time. Edit, approve, reject, rewrite. Nothing is automated. Nothing goes out without the coach's action.
- Approved content reaches the client. Reflection prompts, session continuity notes, accountability check-ins - all carrying the coach's voice because the coach shaped them.
- Between sessions. Coach-approved nudges and prompts keep the client connected to the work. The coach monitors engagement and adjusts. The client experiences continuity. The coach doesn't spend Sunday afternoon doing it manually.
"I'm pretty amazed with both the speed and the accuracy." - Chris Perkins, leadership coach, after his first session. But it was before his second session that something shifted. The preparation brief drew on everything from the first session. "I was blown away with what it came back with. I'm sold."
Session prep that used to take 30 minutes takes five. Notes written from the actual conversation, not from memory. Patterns that might take months to spot surface within sessions.
What does the coach control?
Everything. This is non-negotiable. Specifically:
- What the AI can generate - the coach chooses which features to activate for each client
- What the client sees - every output goes through coach review before delivery
- The coaching framework - the AI adapts to the coach's methodology, not the other way around
- Data access - the coach decides what data is processed and how long it is retained
- Client consent - the coach manages the consent process with their clients
- Override at any point - pause, adjust, or stop AI features for any client at any time
This is the difference between a tool that works for you and a platform that replaces you.
The robo-advisor lesson
Financial advisors went through this exact transition between 2015 and 2020. Robo-advisors like Betterment and Wealthfront launched. The industry split into two camps: "my clients want a human relationship" and "technology will replace us all."
Both were wrong. What actually happened: advisors who understood the technology and used it to augment their practice thrived. They automated portfolio rebalancing and spent more time on the relationship - estate planning, life transitions, the conversations that matter. Their practices grew.
The advisors who said "my clients will always want a human" were correct. But the ones who also said "therefore I don't need to understand the technology" lost clients to advisors who offered both.
Coaching is following the same curve. 37% of coaches view AI as a threat (ICF Global Coaching Study, 2025). 65% use no AI tools for coaching at all. Meanwhile, their clients are already using ChatGPT.
The coaches who figure out how to be in the loop - supervising AI rather than ignoring it or being replaced by it - will own the next decade of the profession.
Try it free with your first client.
Full access, no credit card. Or join the free webinar first - 2 x 60 minutes, no pitch.