Weekly Check-In Triage: 5 Decisions That Improve AI Coaching

Why the best check-in systems sort by decision quality, not message volume.

The strongest finding in the Kahunas check-in corpus is blunt: conversation sequencing and check-in handling dominated the data, with 800 topic hits in the active ingestion report, ahead of nutrition periodization, PED education, and training splits. The mechanism is simple triage: fewer, better decisions made on the right signal beat more “coaching” that only adds noise. In practice, the weekly check-in is not a diary review; it is a decision filter. If an AI coach cannot sort update types into the right action bucket, it will be busy, polite, and wrong.

The real job of a weekly check-in

A weekly check-in should answer a narrow question: what changed enough to warrant a different action this week?

That matters because most physique coaching is path dependent. One bad meal, one underreported fatigue spike, or one misleading photo can send the entire conversation in the wrong direction. The coach’s job is to separate three things that often get blurred together:

Signal — the change that actually changes the plan.
Noise — data that feels important but does not alter the week.
Non-urgent risk — something worth watching, but not worth overreacting to today.

AI coaching is useful only if it improves that sorting.

What the corpus says good triage looks like

The clearest example is contest-prep interpretation. In the peaking corpus, Justin repeatedly rejects the false precision of body-fat estimates: calipers are wrong, hydrostatic is closest, and only dissection gives a true number. The practical move is not to chase a number at all. If the athlete is already lean and the photos show the same thing the number would show, the number adds no information.

That same logic applies to weekly check-ins. If the update does not change the decision, it is not a decision input.

A useful AI check-in system should therefore ask, in order:

Did bodyweight trend move outside the expected range?
Did performance, fullness, or recovery shift enough to matter?
Did adherence change in a way that affects interpretation?
Did digestion, stress, sleep, or schedule create a new constraint?
Is this a real deviation, or the same pattern in a different wrapper?

The point is not to gather more data. The point is to sort what the data means.

A triage hierarchy coaches can actually use

1) First sort by reversibility

The most important question is whether the issue is reversible inside the current week.

If a missed meal, travel day, or bad weekend explains the update, the first action is usually to normalize execution and wait for the next data point.
If the athlete is drifting despite good adherence, the plan needs a real adjustment.

This is where AI often fails: it treats every negative change as requiring a novel intervention. In reality, many check-ins are just execution problems wearing the costume of programming problems.

2) Then sort by magnitude

Justin’s nutrition case work makes a useful point about long-term compounding. He notes that a 5% difference over a year may be invisible in the moment but meaningful over time. That is a good reminder for coaches, but it cuts both ways: most weekly fluctuations are too small to justify action.

If the athlete is within the normal swing of the current phase, don’t force a change just to “do coaching.”

A competent check-in triage system needs thresholds. Not rigid dogma, but enough structure to prevent emotional overreaction to routine variance.

3) Then sort by phase

A high-day carb-up, a medium-day execution check, and a final-week contest-prep adjustment are not the same kind of problem.

The nutrition corpus is very clear that context governs interpretation. On high days, fruit becomes a minor issue because carbs are already high and insulin is elevated all day; the main objective is food throughput. On medium days, more complex carb sources matter a bit more. Pre- and post-workout, fruit is fine up to about half the carbs in those meals. The same food can be harmless in one phase and suboptimal in another.

That is the broader coaching lesson: a check-in is only meaningful relative to the current phase. AI that ignores phase will give generic answers that are technically true and operationally useless.

4) Then sort by decision consequence

The best check-in questions are the ones whose answers can change the next seven days.

Examples:

A small drop in pump and fullness during peak week may be information; it may also be noise.
A clear change in digestion during a critical run-in period may warrant a change in meal structure.
Worsening blood glucose in an enhanced athlete is not a cosmetic concern; it changes the risk-benefit discussion around the current approach.

The PED corpus is useful here because it shows the proper order of operations: if a compound helps bodybuilding but worsens the blood sugar problem, the coach must state the tradeoff directly. That is decision quality. Not all useful coaching is comfortable coaching.

5) Only then sort by urgency

Urgency is where AI can be most misleading. It tends to elevate whatever is most recent, most dramatic, or easiest to summarize. But check-in triage should prioritize what is both important and time-sensitive.

In the peaking material, Justin’s stance on body-fat estimation is a model here: a more exact number is not necessarily a more urgent number. If the athlete is clearly lean and the real issue is filling out and thinning the skin over the final month, then the urgent task is not measurement; it is preparing the look.

A weekly check-in should do the same thing. It should tell the coach what must be acted on now, what can wait, and what should be ignored until the next reporting cycle.

Where AI helps, and where it hurts

AI is genuinely useful in one narrow way: it can standardize the first pass.

That means it can:

extract bodyweight trend, adherence, hunger, sleep, performance, and digestion into a consistent format;
compare the current check-in to prior weeks;
flag mismatches between reported effort and observed trend;
remind the coach when the issue is phase-specific rather than global.

But AI hurts when it tries to replace judgment with summary.

A summary is not a decision.

If an AI model sees “lower energy, flatter look, mild hunger, travel, missed meal” and outputs a generic suggestion, it has not triaged anything. It has just rearranged words. The useful output is a ranked list of likely causes and the one or two actions that materially change the next week.

That is the standard coaches should demand.

A practical weekly check-in template

Here is a triage-first structure that fits the evidence and keeps the coach from drowning in details:

1. Status

Current phase
Bodyweight trend
Adherence score or brief adherence note
One-line performance note

2. Constraint

Sleep
Travel
Digestion
Schedule disruption
Pain or fatigue that affects training execution

3. Interpretation

Is the trend expected for this phase?
Is the issue likely execution, recovery, or programming?
Does this change the plan now, or only the watch list?

4. Action

Keep plan
Tighten execution
Adjust food structure
Adjust training stress
Reassess next check-in

That last line is the real output. Not “here’s what happened,” but “here’s what we do next.”

The falsifiable thesis

If an AI coaching system cannot improve weekly triage, it is not improving coaching. It is only increasing throughput.

That thesis is testable. Compare two systems over a block:

System A generates richer summaries.
System B produces fewer but better action calls.

The better system is the one that makes fewer unnecessary changes, catches meaningful deviations sooner, and leaves fewer ambiguous check-ins unresolved by the end of the week. In serious coaching, that is decision quality. Everything else is decoration.

Sources Used

modules/03-knowledge/kahunas-coaching-deep-nutrition.md
modules/03-knowledge/kahunas-coaching-deep-contest-prep-peaking.md
modules/03-knowledge/kahunas-coaching-deep-peds.md
ingestion-report.md
wiki/troponiniq-kb.md