The Triathlete Blueprint Newsletter #117-Is AI a Threat or a Tool for Coaches?

Yan Busset
Jul 26
9 min read

Read time: 5min.

By Coach Yan Busset

Coaching in the Age of AI

Only two or three years ago, when AI first hit the scene, it felt like a mind-blowing breakthrough. What seemed like science fiction was suddenly real and evolving at an incredible pace. And you know its here to stay and just the beginning when in a so short time its a part of our daily environment. It's everywhere. Its presence is fascinating, unsettling, intriguing, kind of scary… but above all, it’s inevitable.

And yet, I rarely see coaches in the endurance world openly talking about it. As if it were somehow taboo.

Personally, I believe that not using AI today, as a coach, is is borderline malpractice.My responsibility as a coach is to use every valid tool available to help my athletes reach their goals. And if that’s means putting our ego aside and doing works best so that they can achieve their goals.

That said, let’s be clear: AI is not the alpha and the omega. Not yet. It’s far from perfect. But this article isn’t meant to be a manual. I’m not here to explain how to use AI in your coaching. This is something else: a reflection on where we are right now and where this could all be going in the months and years ahead.

What Kind of AI Are We Talking About?

Before we can talk about impact, we need to clarify what we’re actually referring to. Because “AI” is a big umbrella term—and not all forms of AI are the same.

Let’s break it down into three broad categories.

Analytical AI: This one is actually not so new. It’s what we’ve been doing for decades with data analysis and basic algorithms. It reads numbers, like heart rate, pace, training load, HRV, and shows patterns. The AI element here is simply faster processing, automation, and the ability to cross more variables at once. But let’s be honest: there’s nothing revolutionary in concept. It’s just stats, done faster and more smoothly.
Predictive AI : This is where things get more interesting. Predictive AI tries to forecast outcomes based on your past data. For example, based on your last 6 weeks of training and sleep, it might suggest your risk of injury is increasing, or that you're likely to underperform next week. Technically, it uses machine learning models, not language models. These are trained on structured data (numbers, timelines, physiological inputs) and build a statistical model to estimate future events. It’s useful… but also full of assumptions.
Generative AI: This is the buzzy one. It’s where tools like ChatGPT, Claude, Gemini… belong. These are LLMs—Large Language Models. They don’t forecast. They generate.
You ask a question, and they answer. You give them prompts, they write your emails, summaries, used in Chatbots.
Generative AI (Media-Based): This includes tools like DALL·E, Midjourney, or Stable Diffusion. These models don’t generate text, but images or visual content based on a prompt. They are not language models, but diffusion models, designed to build visuals from noise. They won’t generate a training plan, but they can help illustrate drills, create custom infographics, or visualise race gear. Used well, they’re a great support for teaching and communicating more clearly.

But let’s be clear these generative AI what they do is generate sequences of words that statistically make sense based on what they’ve seen before.

Generated, Not Creative

And here’s something we really need to underline: all of this especially generative AI is impressive, but it’s not creative. It can generate, yes. But it doesn’t create in the way a human does. It doesn’t invent from thin air. It doesn’t connect unrelated dots through instinct. It doesn’t innovate.

As Luc Julia, co-creator of Siri, says in his book “IA génératives, pas créatives – L’intelligence artificielle n’existe (toujours) pas”, generative AI is not creative AI. It’s good at producing content that resembles creativity. But there is no intent. No emotion. No lived experience behind it.

And that brings us to another essential limitation: AI is only as good as the data it’s trained on. If the model was fed low-quality content, the output will reflect that. And when it comes to training metrics and athlete input, it’s the same rule.

Garbage in, garbage out.

If your athlete logs false or incomplete data, wrong zones, missing recovery, sessions done differently than recorded, then any AI-based recommendation built on that becomes flawed. The same goes for poor-quality datasets used in training the models themselves.

AI might sound human. It might look like it's learning. And yes, the line between human and machine is getting thinner with each new version. But for now, what remains clearly human is the creative spark, the emotional read, and the ability to create from instinct, not just prediction.

Limits?

One thing that needs to be said clearly is this: the truth of today is not the truth of tomorrow. Especially in the world of AI. What feels accurate or advanced right now might be completely outdated in six months, or even irrelevant in a year. The pace of evolution is extreme.

That said, it’s still worth pointing out some current limitations. One of the risks people rarely mention is what I’d call the snake eating its own tail. AI models now generate so much content online, and those same models are trained on the internet. So they start learning from content that was actually written by… AI. It’s like making a photocopy of a photocopy of a photocopy. The quality declines. There’s a kind of data dilution happening, a loss of sharpness in the source.

And when I look at the current coaching apps on the market, I haven’t yet found one that, that would satisfy me as an athlete or truly impress me as a coach.

Some are too slow. Others are too complex. And many don’t feel like real AI. They feel more like a pre-filled database or a chatbot that pushes out prefab answers from a script. There's little depth.

So for now, in terms of what I do professionally, I don’t see this as direct competition. Not yet. Maybe in 5-10years years, people won’t know if it’s really me behind the screen or an AI version of me. Maybe. But right now, and in the next few years, I don’t see that happening.

Also, we have to be honest: people who don’t like human interaction already avoid coaches. They’re not looking for feedback or dialogue. They’re already using generic training plans, free ones or paid ones. They’re not the ones knocking at my door, at least not yet, they tend to come after they see the limitation of self coaching.

What AI still struggles with is human bias. When people fill out forms or questionnaires (whether for a coach or for an AI) they often present an idealised version of themselves. Lot of them are “ pleasers” . They say they eat cleaner than they actually do, or fail to ignore red flags and try to carry on with training when they should be resting. And the AI has no way to detect that, if the athlete does not log all or don´t have wearables that would track their training readiness.

But as a coach, I often read between the lines. I see things in the patterns, in the inconsistencies, in the tone. I know when someone is pushing too hard or avoiding the real issue. It takes years of experience—and a good dose of diplomacy and psychology—to handle that correctly.

And for now, AI still has a long way to go before it can replicate that kind of human understanding.

Using AI Daily, But Still Waiting for the Right Coaching App

I use AI every day. It has already replaced Google for me in many situations. For quick answers, summaries, comparisons, it’s incredibly efficient. It saves time and provides clarity in a way that search engines simply can’t match anymore.

Sometimes, I even try to use it to generate training sessions. For simple things—standard running intervals or basic cycling workouts, it does a decent job. But when it comes to more complex, nuanced sessions, like swimming drills that require technical progression or adaptation to a specific athlete's needs, it often misses the point completely. That level of subtlety and creativity just isn’t there yet.

Of course, it’s going to improve. No doubt about that. But I still haven’t seen an app, even today, that would really satisfy me if I were the athlete.

Because for me, coaching is about human connection. It’s why I do this job. I want to be next to my athletes. I want to watch them evolve. I want to be there after training, when we grab a coffee and chat. I run in-person groups, and honestly, one of the best tools any athlete can have isn’t an app, it’s training buddies. Real people to train with. To talk with. To share the journey.

That’s something AI can’t replicate.

Spending time with athletes in the real world also helps me learn so much informally. I can hear when someone says “I’m fine,” but I know they’re not. I can feel hesitation, or notice when motivation is dipping, just by observing. You don’t get that through a screen. AI can’t pick up on it. Not yet.

That said, I see massive potential in AI, as a tool for coaches. Especially in processing years of training data. Trends. Red flags. Things that might take hours for a human to notice can be flagged instantly by an algorithm. And that can help remove subjectivity from both sides. It gives the coach better insight and the athlete better objectivity.

But as always, the coach remains the filter. AI can generate suggestions, patterns, and plans, but it can’t evaluate their relevance in a real-world context. That’s where coaching experience comes in. The coach has a kind of built-in BS detector. When the AI outputs something that looks good on paper but doesn’t make sense in practice, the coach sees it instantly. The athlete might not notice. But the coach will. And that’s where the difference is made.

So no, I’m not afraid of AI. I think the best coaches are the ones who adapt. Who stay curious. Who use the tools of their time.

And we live in a time where AI is part of the landscape. There’s no point in asking whether we should use it or not. We have to. But it has to remain a tool, no matter how powerful it is. That’s the key.

I know this will shift many things in society, far beyond coaching. That’s another conversation. But in our field, I’m still waiting for the real coaching apps to emerge. The ones that truly combine the power of AI with the wisdom of human experience. The ones that can blend automation and insight. Structure and empathy.

Nothing on the market does that yet. But I believe it will come. And when it does, it will be a game-changer.

So no, AI isn’t going to replace all coaches tomorrow. Not by itself. The real short terms risk isn’t the machine, it’s the person who knows how to use it better than you.

Because someone who understands coaching and knows how to leverage AI will simply work faster, smarter, and scale further. And that’s where the difference will be made.

But here’s an interesting paradox: the more artificial everything becomes, the more value human presence might gain. Human connection, creativity, empathy, these may become rarer, and therefore more precious. Yes, we have our flaws: subjectivity, inconsistency, emotion. But maybe AI can help smooth out some of those weaknesses. Maybe it can give us space to push our creativity further, by taking care of the tasks that limit us.

That’s the optimistic version. A future where AI doesn’t replace the human but enhances it.

Of course, there’s a more dystopian scenario too. Something like Skynet taking over the world. But let’s be honest, we’ve probably got a few good years left before that happens.

In the meantime, the worst thing we can do is ignore what’s already here. Pretending AI doesn’t exist, or refusing to engage with it, is not a strategy.

AI is a powerful tool. One that can help us coach better, see more clearly, and make smarter decisions, if we use it wisely, and without losing sight of what makes our profession human.

Because in the end, coaching is about people. And that will always be our edge.

Check out my Youtube Channel: