AI Coaching at 180 Clients: the Judgment Stack

AI Coaching at 180 Clients: the Judgment Stack

Coach leverage comes from using AI to widen the roster, standardize the repeatable, and keep human judgment on the decisions that actually matter.

Justin Harris reported that retatrutide at 2 mg produced “no appetite whatsoever” after a single Friday dose, while also making him more fatigued than normal; the mechanism is simple appetite suppression. That is useful evidence for a larger point: AI in fitness coaching is only worth adopting when it increases coach leverage without outsourcing judgment, and the sharp thesis is that AI should scale routine monitoring while the coach keeps authority over load, food, escalation, and tradeoffs.

The hype version of AI coaching says the software becomes the coach. The better version says the coach gets a bigger roster, tighter attention on more athletes, and more time for decisions that require context. Those are not the same thing. A coach does not need an algorithm to know that more food, more hunger, more fatigue, or more insulin sensitivity are not abstract data points; they are decision inputs that change the next move. The value of AI is that it can collect, sort, and surface those inputs at scale. The value of the coach is that he can decide what they mean.

That distinction matters because bodybuilding coaching is not a single problem. It is a pile of repeatable problems and a smaller set of judgment problems. The repeatable side includes check-ins, trend tracking, adherence prompts, and flagging changes in body weight, appetite, fatigue, and training performance. The judgment side includes whether to lean out while appetite suppression is strong, whether to pause or reduce a tool when the tradeoff changes, whether a client is merely stable or is starting to drift, and whether the plan is actually serving the athlete’s phase.

Justin’s retatrutide example is useful precisely because it is not dramatic in the internet sense. It is practical. A 2 mg dose changed appetite fast enough that the response was obvious in real life, not theoretical. He also noticed fatigue and immediately framed the tradeoff in coaching terms: if the athlete is entering a period where food needs to go up, the dose may need to come down or pause; if body composition is moving the right way, it may be worth leaning out while the effect is easy to use. That is the level of decision-making AI should support, not replace.

If you are coaching five athletes, you can often keep these tradeoffs in your head. If you are coaching fifty, you cannot do that reliably without systems. That is where AI earns its keep. It can ingest the same kind of real-world signal over and over: appetite, fatigue, rate of weight change, training quality, meal timing issues, and subjective compliance. It can summarize the pattern, highlight the outliers, and make the coach look at the right athlete first. That is coach leverage: not “automation” in the brochure sense, but attention allocation.

The best use case is not letting AI decide. It is letting AI compress the time between signal and review.

That matters because training and diet are full of thresholds that are visible only when the data is organized over time. A check-in today does not mean much by itself. Three check-ins, one training log, and a pattern of rising hunger or falling output may mean something. A coach with good systems can spot that quickly. A coach with AI can spot it across far more people, but only if the system is built to expose trends rather than bury them in noise.

This is also why the “AI as a replacement coach” pitch keeps missing the mark. Good coaching is not just answering questions; it is assigning weight to the right information. Two athletes can report the same symptom and need different responses because their phase, history, compliance, and targets differ. No model can infer those priorities cleanly from a checklist alone. A coach can, because judgment lives in the context that does not fit neatly into a form.

TroponinIQ sits in the useful middle: an AI coaching layer on top of a real coaching philosophy. The platform exists in a system associated with 100+ pro athletes and 1,000+ transformations, but the point is not the bragging right. The point is that a large body of coaching experience can be made more available without pretending that every answer is generic. That is what serious coaches should want from AI: a way to scale the best parts of the process without flattening the differences between athletes.

There is a second leverage point that matters even more than monitoring: standardization. AI can help create a more consistent intake language, more consistent progress summaries, and more consistent flagging of problems. When the same athlete reports “fatigue,” one coach may hear “normal training stress” while another hears “pull back now.” A system can force better structure: What changed? Since when? How severe? What else changed with it? That discipline does not make the coach obsolete. It makes the coach better informed.

The limitation is obvious and important. Standardization is not wisdom. If the software turns coaching into a box-ticking exercise, the athlete loses. If it is used to generate too many alerts, the coach loses. If it is used to make decisions from averages instead of from the actual person in front of you, everyone loses. The best coaches will use AI the way a good assistant is used in any high-skill profession: to prepare the work, not to own it.

So the practical model is straightforward. Use AI for the boring but necessary layers: tracking, summaries, trend detection, routing, and reminders. Use the coach for phase decisions, appetite tradeoffs, escalation, and anything that depends on how one variable changes the meaning of another. The more athletes you roster, the more valuable that separation becomes. The goal is not to create a machine that coaches. The goal is to create a coaching operation that can think clearly at scale.

That is the real coach leverage story in AI fitness coaching: bigger roster, better filtering, same judgment. Not replacement. Not magic. Just a better division of labor.

Sources Used

• raw/_consumed/2026-05-26/troponiniq_kb.md
• raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/clients/rory_lazowski___members-c5balaovjbdoeefqmfuqdhh2tbpmfdu16lnf0tnrtmw.json
• raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/transcripts/rory_lazowski___members-c5balaovjbdoeefqmfuqdhh2tbpmfdu16lnf0tnrtmw.md
• raw/Justin_TT1.txt