I decided to try and build an AI Coach. When it comes to trying out and learning about all the new AI products tools and toys, one statement stood out for me: “Despite your fear, you should try and see if you can replace our tasks and roles with AI”. Since I am constantly learning about and trying to build with AI and also do coaching alongside my main role, I thought building an AI coach would be an interesting challenge.
So I started working on it. And the more time I spent, the more I realised the obvious limitations that an AI Coach has, at least for now. Here’s what came up so far.
An AI Coach Can’t Sense The Emotions or Tone In Voice
Communication in general, but even more so coaching, doesn’t just live in the words people say. It lives in everything that comes with those words. The shift in tone or volume or talking speed, the change in energy, the switch in body language, the unconscious gestures, the fidgeting with a pencil or hair.
And a text-based AI coach misses aaaaaaall of that. Even voice-enabled AI tools are doing pattern recognition but might miss the meaning of all signals put together. The AI counterparts are not truly listening or not able to listen at all in the way a coach does, to pick up all communication signals and listen to the whole person, not just the written words.
When it comes to communication, 93% of it is non-verbal. around 50% is attributed to body language and around 40% to tone of voice. So an AI Coach seems to only deliver 7% of the maximum potential performance.
An AI Coach Doesn’t Know What To Do With Silence
I actually wrote a whole article recently about the importance of silence in coaching. Silence in coaching is often where the real work is happening. When a coachee goes quiet because a question or insight just unlocked something for them – that is one of the most valuable moments in a coaching session.
An experienced coach recognises that silence and holds it. holds the space. Doesn’t rush the moment or fill the gap. An AI Coach? In case of pure text based coaches, they won’t even be able to detect the silence to begin with. Because text based communication doesn’t show that unless the coachee writes that they are reflecting or silent, which of course, won’t happen. And in case we have a conversational agent that would be able to listen in, that might also be constrained just by the partterns that it detects. But even then, it will only perceive silence in isolation without seeing what the coachee is doing. It has no way to evaluate: “Is this good silence or bad silence? Is this where I wait or where I jump in and ask something?” Which brings me to the next limiting factor…
An AI Coach Can’t See You
I touched upon this already in the first 2 issues that most communication is non-verbal. And over 50% of that is body language. So this coach I am trying to build is limited to only 50% of the communiation from the get go. And the worst part about this is that there are times when the words say one thing but the body language actually says something completely different. That’s where it’s worth asking and challenging. That’s when the session starts becoming even more valuable. And sadly, the ai coach (at least for now) simply doesn’t have access to that.
An AI Coach Can’t Create The Safe Space
Coaching needs trust and safety. Not just intellectual or professional trust. Or trust that the tool is smart or secure enough. Psychological safety. The feeling or sense that you can say the thing you haven’t said anywhere else. That it’s worth sharing in this space. That you won’t be judged. That sharing or opening up in this space with your coach is valuable and has meaning.
That kind of trust is build slowly, through presence, consistency, the feeling that you are properly listened to. Due to everything I’ve mentioned and encountered above, it’s kind of difficult right now for an AI coach to earn it to the same degree. And without that trust and psychological safety, the risk is that the coachee won’t go deep. They risk never going beyond the surface on the topics and challenges they bring to the table.
Ok, So What Does This Actually Mean?
Well, it means a few things.
For the coachee: without the ability to listen fully to body language, tone, feelings, silence or to gain trust and create a safe space, there’s a very real chance that the session will only produce a surface – level conversation rather than genuine insights and changes in perspective. The value they should have gotten? Only partially delivered. This is definitely not the outcome that quality coaching aims to achieve.
For me as the coach (and builder): based on what I’ve built and experienced so far, genuine human coaching is still hard to replicate. I am still trying out to see how it might add value when it comes to areas like reflection or accountability and it definitely has its uses for structured thinking or partial preparation befor ea session.
But a supplement does not a subsitute make. Have you tried building an AI coach or used one? What has your experience been so far? Would love to read your thoughts in the comments!
Looking to spark change in your journey? Here’s how I can support you:
- Product Management & Leadership Coaching – I help leaders grow their confidence, lead despite uncertainty, focus on delivering value and create real impact for their teams, users and company.
- Outcome Thinking & OKRs – Backed by hands-on experience in both using and rolling out OKRs across multiple teams and companies, I help you roll out, improve or fix your OKRs or OKR system and make changes and improvements that stick beyond the start of the cycle.
- Facilitation With a Twist – I create, facilitate and moderate workshops that energize the room and foster stronger engagement and learning – all through my crazy, fun and playful approach.
- GenAI for Product Teams – Practical, hands-on workshops that give teams and leaders the skills to harness AI and GenAI for their everyday goals and way of working, but also the time to reflect as a group on the impact that this AI wave brings.