Weekly Check-In Triage and 3 Decision Rules

Weekly Check-In Triage and 3 Decision Rules

Why the best AI coaching use case is not motivation, but classification under uncertainty.

The strongest pattern in the check-in corpus is simple: coaches made better decisions when they treated weekly updates as triage, not storytelling. In Justin Harris’s notes, the actionable signal is usually a concrete change in weight, digestion, photos, readiness, or adherence, with the mechanism named in plain language: fluid balance and GI transit. That leads to a falsifiable thesis for AI coaching: if your weekly check-in system cannot classify risk and prioritize action from a handful of stable inputs, it is not improving decision quality — it is just producing more text.

Weekly check-ins are decision systems, not summaries

A lot of coaching software tries to make the check-in feel comprehensive. More fields, more sliders, more sentiment, more commentary. The problem is that decision quality does not rise with questionnaire length if the extra input does not change the next action.

The corpus shows the opposite pattern. In one prep message, a coach reacts to a “great week of progress” but immediately anchors on the practical issues that matter next: a big weight drop, constipation, and distension. The reasoning is mechanistic and narrow: travel reduced water, the gut backed up, and distention rose. The response is not “tell me more.” It is to classify the problem as likely GI and hydration related, then adjust the plan accordingly.

That is triage. Not diagnosis, not narrative, not theory. Just sorting the signal that changes the next week.

For AI coaching, that matters because most weekly reports mix three different categories:

1. Trend data — bodyweight, photos, performance, schedule.
2. Noise — travel, stress, food timing, subjective mood.
3. Decision triggers — a change that warrants action now.

A good check-in system does not treat all three as equal. It surfaces the trigger, tags the likely cause, and tells the coach what to review first.

The mechanism is classification under load

In practice, the bottleneck is not data collection. Coaches already have plenty of that. The bottleneck is deciding what the data means fast enough to keep the plan on track.

That is why the most useful AI layer is not “chat about your week.” It is “sort this week into a small number of action buckets.”

The corpus gives several examples of the bucket logic:

• GI / distension bucket: weight moved, but constipation and abdominal distension appeared earlier than usual. The likely driver named in the message is travel and reduced water intake. The response is to treat the gut as the issue, not the physique.
• Readiness / fatigue bucket: coaches repeatedly lean on a small set of stable markers rather than vague reports. If the data says the athlete is in a good position at the current body fat and dose, the decision is restraint, not escalation.
• Timeline bucket: one note explicitly says the coach dislikes prepping hard for longer than 16 weeks and prefers not to push too much until about 20 weeks out. That is a decision rule about timing, not a motivational slogan.

This is what AI should imitate: not eloquence, but prioritization.

Why weekly triage beats weekly overreaction

Bad weekly check-ins create one of two errors.

First, they cause false escalation. A coach sees a rough check-in and changes too much too soon. That usually happens when every symptom is treated as a sign that the whole plan is failing.

Second, they cause false reassurance. The athlete reports progress, but the update hides a developing issue because the system never asked the right follow-up questions. A coach who only reads averages can miss the part that actually matters.

The corpus points to a more disciplined model: use the check-in to answer three questions, in order.

1) Is there a real trend change?

Weight drop, photos, and adherence are the first pass. If those are moving in the expected direction, do not panic because a single subjective complaint appeared.

2) Is the complaint local or global?

Constipation and distension are local. They point to digestion, hydration, travel, and food handling. They are not automatically proof that the entire prep is off track.

3) What is the smallest useful action?

If the issue is backed-up gut and low water, the smallest useful action is a targeted adjustment, not a full-program rewrite.

That sequencing is where AI can help. A model that pre-sorts the week into “trend intact,” “GI issue,” “recovery issue,” or “timeline mismatch” gives the coach a cleaner starting point. The coach still decides. The software just reduces search time.

What the system should ask every week

A triage-oriented check-in does not need 40 questions. It needs a short, high-yield set that maps cleanly to decisions:

• Bodyweight trend: Is it moving as expected?
• Photos: Is the visual change consistent with the scale change?
• Digestion: Any constipation, distension, or unusual fullness?
• Hydration / travel: Any recent disruption in water intake, travel, or routine?
• Training performance: Is output stable enough to support the current phase?
• Adherence: Did the athlete actually execute the plan?
• Timeline position: Are we early, mid, or late relative to the goal date?

Those seven inputs are enough to triage most weekly coaching situations seen in the corpus. More fields may be useful later, but they should not block the first decision.

Decision quality improves when the system is narrow

There is a reason experienced coaches sound repetitive when they are being effective. They are not searching for a novel explanation every week. They are checking whether the same few failure modes are present again.

The corpus is full of this kind of practical skepticism. One note explicitly rejects overconfidence: “I don’t have any evidence other than it’s what I think.” That’s not weakness. It is an accurate statement of evidentiary status. A good coach knows the difference between an observed pattern and a firm rule.

That attitude is exactly what AI check-in triage needs. The model should not pretend to know more than the data supports. It should say:

• This looks like a GI / hydration issue.
• This looks like a trend on track.
• This looks like a timing issue, not an urgency issue.
• This looks like missing context; ask one more question.

That is better than a bloated narrative response because it preserves the coach’s attention for the decisions that matter.

Practical standard for coaches

If you use AI in weekly check-ins, judge it by one question: Did it improve the next action?

If the answer is yes, it probably did one of four things well:

• identified the main constraint,
• separated signal from noise,
• reduced unnecessary back-and-forth,
• or escalated the right issue first.

If the answer is no, the system likely produced commentary instead of triage.

That gives you a simple standard. The best AI coaching tool is not the one that sounds most human. It is the one that helps a coach classify the week correctly, fast, and with fewer wrong turns. In weekly check-ins, decision quality comes from narrow buckets, stable inputs, and the discipline to act on the smallest useful change.

Sources Used

• raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w19-24m/clients/david_lamartina___members-tlssnsjthkmnhfqcscszce25acz_vhdm_x2_xdlpx_i.json
• raw/_consumed/2025-05-26/troponin_supplements_kb.md
• raw/_consumed/2026-06-02/_GRAS/gras_strategy_training.md
• raw/_consumed/2026-05-26/troponiniq_kb.md