AI coaching apps are everywhere. They promise personalised training plans, adaptive programming, and data-driven decisions. The pitch is compelling: why pay for a human coach when an algorithm can do it cheaper and faster?
The answer is straightforward. An algorithm does not know you. It knows numbers. And the gap between the two is where most of the meaningful coaching decisions live.
What the Algorithm Actually Does
Most AI training platforms work from population-level models. They estimate your training zones from a single test or, worse, from age and resting heart rate. They generate plans from templates, adjusting volume and intensity based on your stated goals and available hours. Some use machine learning to nudge future sessions based on how you responded to past ones.
This is not personalisation. This is a template with a feedback loop. The underlying model still treats you as a point on a bell curve, not as an individual with a unique metabolic profile.
The Profile Problem
Two athletes can walk into a training block with the same threshold power. One is a glycolytically dominant athlete with a compressed lower zone structure and a narrow gap between their first and second threshold. The other is an aerobically dominant athlete with a broad aerobic base and compressed upper zones.
They need completely different training. The first athlete needs shorter repetitions at threshold to avoid drifting above the intended intensity as metabolic byproducts accumulate. The second needs more work above the second threshold to raise their ceiling. Same number, different physiology, different programme.
No training app can make this distinction. It requires profiling the athlete across multiple durations, examining the shape of their power-duration curve, and understanding the ratios between short and long efforts. A single threshold estimate, however accurate, cannot capture it.
The Context Problem
Your training does not exist in isolation. It sits inside a life. Your body runs a single stress budget, and training is only one withdrawal from it. A week of broken sleep, a project deadline at work, or a house move all reduce your capacity to absorb training stimulus.
A coach sees this. A coach asks how you are before prescribing what to do. A coach adjusts Wednesday's session because Monday's data looked off and Tuesday's message said the kids were up all night.
An app sees your heart rate variability score and your last completed workout. It does not know why your numbers dipped. It cannot distinguish between productive fatigue from a well-executed training block and accumulated life stress that demands a lighter week. It responds to the symptom, not the cause.
The Drift Problem
Olav Bu, the sports scientist behind Norway's Olympic triathlon programme, discovered something instructive about intensity calibration. Even with the most sophisticated data systems in endurance sport, his elite athletes drifted away from optimal training intensities within months when external feedback was reduced. One athlete consistently pushed harder than intended. Another consistently held back too much.
If Olympic-level athletes with years of structured training behind them cannot self-regulate intensity without regular human recalibration, the idea that an app can manage it for an age-group athlete with less training history and fewer reference points is optimistic at best.
Intensity discipline is perishable. It requires ongoing calibration against something external. An app can tell you your target power. It cannot tell you whether that target is still appropriate given what happened in the rest of your life this week.
Tools Are Not Coaches
None of this means training apps are useless. They are tools, and tools have value. A structured plan from an app is better than no structure at all. Tracking your training in software is better than not tracking it.
But tools need someone who understands what the numbers mean. Arild Tveiten, head coach of the Norwegian triathlon programme, has made the point clearly: coaches who learn to observe athletes first and then add technology are stronger than those who start with technology and never develop the observational skill. The technology serves the coaching, not the other way around.
The best training decisions are not made from data alone. They are made from data interpreted through understanding of the individual athlete, their physiology, their life, and the interaction between the two. That interpretation is the coaching. The rest is accounting.
An app can tell you what you did. A coach can tell you what it meant and what to do next.