Every coach who discovers AI coaching hits the same five fears — and none of them are irrational. Here's what four years of doctoral research and a hundred real deployments reveal about what's really behind coaches' resistance to AI, and why it's usually a signal to design better, not a reason to say no.
Every coach, mentor, supervisor or trainer who discovers AI coaching goes through roughly the same five minutes of vertigo. They see the demo, they're impressed — and then something resists. A small voice that says: am I betraying something essential about my profession?
That resistance isn't irrational. It deserves to be heard, not dismissed. But it also deserves to be questioned — because it usually rests on an assumption that doesn't hold up under scrutiny: that a human is always available, that the only alternative to generic AI is silence, and that "tool" means "replacement."
Here are the resistances that come up most often in our conversations with coaches, mentors, supervisors and trainers — and what four years of doctoral research and more than a hundred real-world deployments have to say about them.
It's the most direct fear, and sometimes the hardest to say out loud: if my clients have an AI agent, will they still need me?
It sometimes takes an almost funny, very revealing form. A trainer, while building her own agent with us, half-joked: "Actually the AI is going to replace me, and you'll be in my place." Her own client, in that same conversation, spontaneously replied: "No, I want the real you."
That moment sums it all up. The agent reproduces a method, not a presence. It isn't the coach's double — it's a tool the coach shapes with their own framework, guardrails, red flags, and the ability to recognize the moment to hand back to the human. A professional's value is no longer measured in hours delivered, but in the quality of their methodology and everything the tool can never replicate.
This is the deepest one, and the most legitimate. Something happens in human interaction — felt sense, intuition, reading the room — that could never exist with AI.
That's true. And that's precisely not the agent's job. What a coach picks up from body language, silences, the micro-resonance between two humans in a room — that's exactly what stays reserved for human sessions. The agent handles frequency and daily practice; the human keeps the relational depth. These aren't two competing offers — they're two different timescales.
The question that really moves the conversation forward isn't "does AI replace the human?" but "what does someone do today, at 11pm, when no human is available?" For most people, the answer is: nothing, or unfiltered generic ChatGPT. An AI shaped by a human expert isn't competing with the coaching session — it's competing with silence and improvisation.
Another, more philosophical kind of resistance: if I give my clients an external tool to re-center themselves, am I not preventing them from learning to do it alone?
That was, quite precisely, the subject of the doctoral thesis that shaped Magif.ai's design. Research shows that interacting with an AI coaching agent designed by an expert triggers the same cognitive and affective self-regulation mechanisms as a human session. Nothing is outsourced: the person does the re-centering work themselves, in their own words, in their own situation. That work becomes internalized through practice — exactly like any skill first learned with an outside resource: a trainer, a book, a coach.
This one is very concrete, and very common among non-technical coaches: this is too complicated for me, I'm not comfortable with tech.
A trainer told us bluntly during a demo: "Everything you're saying is fascinating, but to me it's Greek — it's a fog." An hour later, without ever touching a single configuration line on her own, she had a structured agent, tested, scored 9 out of 10. The barrier wasn't technical — it was in what "building an AI agent" meant in her head.
Then comes the content question: if I put my method into the agent, could it get copied, or hacked?
That's a legitimate concern — one coach even shared the story of a consulting firm whose internal AI had been hacked in two minutes by security researchers. Two things help put it in perspective. First, most coaching content already looks remarkably similar from one practice to the next — the real intellectual property isn't in the text, it's in how you execute it with a client, something AI can never reproduce. Second, European hosting and a GDPR framework reduce the risk, without ever claiming a zero risk that doesn't exist anywhere online.
None of these fears should be waved away. They're signals — invitations to design AI correctly, not reasons to close the door. The difference between an AI that develops humanity and one that impoverishes it is never in the technology itself: it's in the design. An AI shaped by an expert, with a methodology, clear red flags and clear boundaries, doesn't produce the same experience as generic ChatGPT with no guardrails.
So the real question isn't choosing between the human and the AI. It's asking what the person with no one to talk to is doing right now — and deciding whether to leave them with silence, a generic chatbot, or a tool built with the expertise of someone who's spent their career understanding what helps people grow.
Dr. Olivier Malafronte is a coach and AI coaching pioneer since 2015, associated researcher and teacher at Université Côte d'Azur (IAE Nice), co-author of the ICF Global AI Coaching Standards, and co-founder of Magif.ai.
