The Coaching Conversation Has a New Voice
Sixteen years coaching runners teaches you to read patterns across time. What changes when AI can do that too?
Sixteen years of coaching runners teaches you to read patterns across time.
Not just “you ran well this week” but “you always tighten up in week three of a build, you’ve done it six times, here’s what happened after each one.” That longitudinal memory is one of the things that separates a good coaching relationship from a good training plan. Plans are static; coaches hold context.
For most of my coaching career, that memory lived in my head, in spreadsheets, in the frantic scrolling back through years of messages when someone said “I don’t know, I just feel like this is worse than before.” I built systems to help. But the cognitive load of tracking 77 athletes concurrently — at the peak of Mile Pursuit — is real. Things get missed. Context gets dropped.
Then I started using AI in my coaching process, and something unexpected happened.
Not “the AI found insights I couldn’t.” That’s not quite it. More that the AI could hold a different kind of attention than I could. I’d feed it six months of training data and ask it to look for the patterns I had a hunch about. Sometimes it confirmed the hunch. Sometimes it found something adjacent. Occasionally it found nothing, which was also useful — my hunch was just a hunch.
What changed wasn’t my coaching intuition. It was the triangulation. The athlete has their felt sense of how training is going. I have my relational read, my pattern recognition across the cohort, my understanding of this person over years. The AI has a kind of forensic recall of the logged data that neither of us has the bandwidth to hold fully.
Three perspectives. Different enough to be useful.
The thing I’m still working out is what this does to the athlete’s agency.
Coaching, at its best, is a permission-giving practice. You’re not telling people what to do — you’re creating conditions where they can hear themselves better and make better choices. Part of that is the coach not having all the answers, not being a source of certainty, being genuinely curious about this particular person in this particular moment.
AI has a different relationship to certainty. It can say “based on everything you’ve logged, your recovery metrics have been declining for three weeks” with a confidence I wouldn’t use, because the data is just one signal among many. Used carelessly, that could actually damage what I’m trying to build.
But used well? I think it deepens the conversation and strengthens authorship. “The data says X — what does your body say, and what do you want to do?” is a better question than either one alone.
I’m still coaching. Still patient, still long-game, still fundamentally permission-giving. The AI hasn’t changed that. The measure is whether the athlete leaves with more agency: a clearer read, a meaningful choice, and ownership of what happens next. Learning when AI helps that — and when to keep it quiet — is part of the craft now.