What 52 Coaches Told Us About AI

91% of coaches told us the same thing: if AI touches my client, I decide what it says. We spoke to 52 of them across 11 countries before writing a line of code. Here is what we learned.

Before we built anything, we spent six months listening. Not focus groups. Not surveys with leading questions. Conversations. One-to-one, usually an hour, sometimes two. Coaches from London to Lagos, Sydney to Stockholm, New York to Nairobi. ICF-credentialed, EMCC-accredited, independent, in-house. Executive coaches. Leadership coaches. Transition specialists. 52 people who do this work every day.

We did not go in with a product to validate. We went in with questions. What is actually hard about your practice? What keeps you up at night? What do you think about AI? And then we listened.

Some of what they told us confirmed what we expected. Some of it surprised us completely. All of it shaped what we built.

What is the biggest problem coaches face right now?

It is the gap between sessions.

85% of the coaches we spoke to confirmed the same problem: their clients want to continue the work between sessions, but the economics of one-to-one delivery make it impossible. A leadership coach charging $500 per session cannot send check-in texts, write reflection prompts, and review progress journals for every client. The maths does not work.

"My best clients are the ones who want more. They email me between sessions, they want to keep the momentum going. And I love that. But I have 14 active clients. If I respond to all of them the way I want to, I have no life." - Executive coach, Singapore

We started calling this the extension problem. Coaches know that the real transformation happens between sessions. The integration, the practice, the moments of insight at 11pm on a Tuesday when something from the session suddenly clicks. But coaches cannot be there for those moments. Not at scale. Not sustainably.

One coach in Melbourne put it simply: "I am selling an hour. But the work is not an hour. The work is the six days in between."

What do coaches actually think about AI?

Not what the headlines say.

The narrative in the coaching world is that coaches are either terrified of AI or blissfully ignoring it. The reality is more nuanced, and more interesting.

We found three distinct groups:

• The cautiously curious (about 50%). They see potential but have deep concerns about control, ethics, and the coaching relationship. They are reading about AI. They have not used it for coaching.
• The openly anxious (about 30%). They worry that AI will commoditise coaching, undercut their fees, or render them irrelevant. Several mentioned BetterUp's AI coach by name. The fear is specific and grounded.
• The early adopters (about 20%). They are already experimenting, mostly with ChatGPT for session prep or note-taking. They see the limitations clearly. They want something purpose-built.

But across all three groups, one finding was consistent.

91% said coach control is non-negotiable

When we asked coaches what would make them comfortable using AI in their practice, 91% gave some version of the same answer: nothing reaches my client without my approval.

Not "I would like to review things." Not "it would be nice to have oversight." The word that kept coming up was non-negotiable.

"If I cannot see it, edit it, and approve it before my client sees it, I am not interested. Full stop. That is not a feature request. That is the entry requirement." - Leadership coach, London

This was the single clearest signal in our research. And it is the reason we built a coach-supervised model, where AI proposes and the coach reviews every output before it reaches the client. Not because it is a nice philosophy. Because 91% of coaches told us it is the only way they would use it.

How many coaches are actually using AI tools?

65% of the coaches we spoke to use no AI tools for coaching at all. Zero. Not for notes, not for session prep, not for client communication. Nothing.

Of the remaining 35%, most were using general-purpose tools in informal ways:

• ChatGPT for brainstorming session questions
• Otter.ai for transcription (then manually writing their own notes)
• Notion AI for organising client records
• Voice memos they never transcribe

Nobody was using a tool built specifically for coaching. Several coaches described cobbling together four or five apps that sort-of-worked. One called it "a Frankenstein tech stack held together by guilt and good intentions."

The gap between what is available and what coaches need is enormous. This is not a market that has been well-served and needs incremental improvement. It is a market that has barely been served at all.

Are clients already using AI without their coaches?

Yes. And this was the finding that worried coaches most.

48% of coaches told us their clients are already using ChatGPT or similar tools between sessions. Clients are typing in their coaching goals and asking for accountability. They are journalling with AI. They are running leadership scenarios through a chatbot.

They are not doing this because they prefer a chatbot to their coach. They are doing it because their coach is not there, and something is.

"One of my clients told me she had been using ChatGPT to practise difficult conversations before her board meeting. She said it was helpful. And I thought - that should have been me. Not because I am possessive, but because I know her. I know the context. ChatGPT does not know that her CFO undermines her in every meeting." - Executive coach, Amsterdam

The urgency here is real. Coaches are not competing with other coaches for the space between sessions. They are competing with whatever tool their clients find first. And right now, that tool has no understanding of the coaching relationship, no ethical framework, and no context.

What surprised us most?

Two things we did not expect.

The emotional weight of ending relationships

We asked coaches about the hardest parts of their practice. We expected them to talk about difficult clients, organisational politics, or the business side of coaching.

What came up repeatedly was the end of engagements. The moment when a coaching programme finishes, when an organisation's budget runs out, or when a client moves on. Several coaches described genuine grief.

"I had a client for two years. We did extraordinary work together. Then his company restructured and the coaching budget was cut. He could not afford to continue privately. We had one final session and that was it. I still think about whether the work held." - Leadership coach, Toronto

This connects directly to the extension problem. Coaches do not just want to extend engagement for revenue (though that matters). They want to know the work continues. They want a way to maintain the developmental relationship without requiring the economics of hourly sessions.

The fear-curiosity paradox

The second surprise: most coaches hold both fear and curiosity about AI simultaneously. These are not opposing camps. They are the same people.

A coach in Berlin told us she was "terrified and fascinated in equal measure." A coach in Johannesburg said he "reads every article about AI coaching with one hand covering his eyes." The analogy that came up most often was watching a car crash in slow motion while also wanting to learn to drive.

This is important because it means coaches are not waiting for permission or persuasion. They are waiting for a model they can trust. When we described coach-supervised AI, where the coach reviews and approves every output, the response was consistently relief. Not excitement. Relief.

What did the numbers look like?

Here is a summary of the key findings from our conversations.

What did we build because of this?

Everything. Every feature decision traces back to these conversations.

Coach-supervised AI. Because 91% told us control is non-negotiable. The AI proposes session notes, reflection prompts, and check-ins. The coach reviews, edits, and approves every output. Nothing reaches the client without the coach's action. This is not a setting you toggle on. It is the architecture.

Session intelligence. Because coaches told us the real value is not transcription, it is the patterns. What themes keep recurring? What language is the client using differently? What did they commit to three sessions ago that they have not mentioned since? These are the observations that make coaching powerful, and they surface faster when an AI is helping you look.

The client extension model. Because 85% confirmed the extension problem. Coaches can now send approved reflection prompts, accountability nudges, and continuity notes between sessions. The client stays connected to the work. The coach stays in control. The relationship deepens without requiring more hours.

The coaching twin. Because coaches told us they want AI that sounds like them, not like a generic chatbot. The system learns the coach's voice, their frameworks, their language. When a client receives a reflection prompt, it feels like it came from their coach. Because functionally, it did.

We did not build a product and then find coaches who liked it. We found coaches, listened to what they needed, and then built the product. That distinction matters.