A coach sits across from a client who has just gone quiet. Something has opened up, and the person does not yet have words for it. The silence lasts ten seconds. Then twenty. The coach does not fill it.
That moment does not need an AI intervention. It needs a human being with the judgement to stay with what is happening.
AI could generate a prompt. It could suggest a follow-up question or offer a tidy interpretation of what the client might be feeling. It might even produce something plausible. But plausibility is not the point. The coach chose not to speak, and that choice carried the moment.
We build AI tools for coaches, so we think about this constantly. The more we build, the clearer it becomes that the most important design decisions are not just about what AI can do, but about what it should not take over.
For us, the line is simple: AI should support coaching in ways that strengthen the coach's presence, judgement, and continuity. It should not displace the human relationship at the centre of the work.
The real question is where AI belongs
There are many ways AI can be useful in coaching. It can help a coach hold onto the thread between sessions, reduce administrative load, surface themes over time, and make it easier to sustain reflection and accountability.
None of that is inherently a problem.
The problem begins when AI stops supporting the coaching process and starts replacing the human responsibilities that make coaching effective in the first place.
Coaching is not a sequence of technically correct interventions. It is a relationship in which two people do difficult work together. The quality of that work depends on trust, timing, judgement, and presence. The hard parts are not inefficiencies to be designed away. They are often the work itself.
So the question is not whether AI can help. It is whether the coach remains clearly in the lead.
Five things AI should not take over
1. The silence
A coach who sits in silence with a client is not doing nothing. They are holding space. They are communicating, without a word, that the client does not need to rush to explain, perform, or move on too quickly.
AI may help a coach reflect afterwards, or help carry themes forward between sessions. But in the live moment, silence is not a gap to be filled automatically. It is often part of the work itself.
2. The relationship itself
The coaching relationship is not just a channel through which insights are delivered. It is the foundation that makes the work possible. Trust is built through attention, consistency, challenge, and care over time.
AI can support that relationship. It can help a coach maintain continuity, reinforce commitments, and extend reflection between sessions in ways shaped by the coach's intent. But it should not replace the coach as the primary source of trust, challenge, and direction. The client should remain clear that the coaching relationship is with a human being.
3. The moment of confrontation
Good coaches sometimes name what the client is avoiding. They notice the pattern, take a risk, and say the difficult thing clearly.
AI can help surface patterns for the coach to consider. But the decision to confront, when to do it, and how to do it belongs to the coach. Those moments depend on judgement, timing, and relational trust, not just pattern detection.
4. The ethical judgement call
Sometimes a client brings something that sits at the edge of the coaching contract. It may point to acute distress. It may require referral. It may raise questions about what coaching can and cannot hold.
AI may assist by surfacing signals or helping the coach document concerns. But ethical responsibility cannot be delegated. The judgement, and the weight of it, stays with the human practitioner.
5. The coach's intuition
Experienced coaches develop a feel for what is happening beneath the surface. They notice shifts in tone, hesitation, contradiction, deflection, and the meaning of what is not being said.
AI can offer another layer of pattern recognition across sessions, notes, and reflections. That can be useful. But it is not a substitute for the coach's lived, relational, embodied sense of the person in front of them.
Where AI can support coaching
AI is most useful where it strengthens the coach's capacity without displacing the relationship at the centre of the work.
It can capture and organise what happened in a session so the coach is less burdened by note-taking and recall. It can help surface themes, commitments, and changes over time so continuity does not depend on memory alone. It can support reflection and accountability between sessions in ways shaped by the coach's approach. It can also help coaches review their own practice and notice patterns they may otherwise miss.
This matters because most coaches want to offer more continuity than time allows. They want to remember the important thread from last month, follow up on the commitment that mattered, and support reflection between sessions without adding unsustainable administrative work.
Used well, AI can help with that. Not by taking over the work of coaching, but by reinforcing the conditions that make good coaching more possible: presence, continuity, intentionality, and follow-through.
Where we draw the line
At Mentri, we use a simple test: does this strengthen the coach's work, or does it start to displace the coach's judgement, presence, or relationship with the client?
If it strengthens the coach, we are interested.
If it weakens the coach's role or makes AI into an autonomous coaching authority, that is a line we do not want to cross.
That principle shapes how we build. We believe AI can play a valuable supporting role in coaching. But its value comes from extending the coach's capacity, not replacing the coach's responsibility.
The best coaching technology does not become the centre of the conversation. It helps the coach stay prepared, present, and connected to the work across time. It supports the relationship without trying to become the relationship.
AI in the hands of coaches, not coaching in the hands of AI.
