Later path · Metabolic Health Diagnostic
This is a bounded field engagement for organizations that already understand the strategy problem and need a hard look at one decision loop where AI is creating friction. It is not the default first step. Most leaders should start with the Strategy Engineer AI agent and QCEA practice, then redesign for metabolic health. The diagnostic is how we sometimes go deep in the field afterward.
When it runs: we follow one consequential loop end to end. You learn what is slowing it down, how confident we are about each finding, and what the next ninety days should contain. The Strategy Engineer agent supports the work; a practitioner remains accountable for what reaches you. You keep everything we produce whether or not you work with us again.
Founding engagement
This is a founding-cohort price, and we are being deliberate about saying so. The method is worked out and the deliverables are specified, but we are early enough that we are still instrumenting our own delivery — measuring how long each phase actually takes rather than quoting numbers we have not verified. You get founding pricing; we get evidence. If you would rather buy something with a decade of case studies behind it, that is a reasonable preference and we are not it yet.
Who buys this
This works when it is bought by someone with direct accountability for a specific workflow, not by a committee evaluating a transformation vendor. Both tiers sit below the discretionary threshold at most organizations, which is deliberate: the point is to skip procurement, not to survive it.
You have spent real money on copilots and agents, and cycle times have not improved. Ticket re-opens are up, and so is the platform bill. Nobody can tell you which of those three facts caused the others.
Your team is drowning in machine-generated alerts. Humans are signing off on findings they did not have time to actually check, and there is no record of what was genuinely verified versus what was waved through.
Intake got faster. Then exceptions started arriving downstream, and the rework is eating the margin the speed was supposed to create. The gain and the cost sit in different teams’ numbers.
AI is pulling roles apart faster than you can put new ones together. The tacit knowledge that made the old roles work is walking out, and the people left are absorbing the difference.
Scope discipline
Ten days is enough because the boundary is narrow, and the boundary being narrow is what makes the findings hold. One decision loop, agreed in writing before we start, with a defined beginning and end. Three examples of what that looks like in practice — yours will be your own.
Agents generate remediation faster than engineers can responsibly verify it, so a verification queue forms behind the automation and the net cycle time gets worse, not better.
Summarization accelerates intake, but non-standard conditions do not survive the summary. The rejections surface downstream, in a different team, against a different metric.
The automated path works until an exception, at which point the real state of the account lives in private messages and has to be reconstructed verbally before anyone can move.
We build a first draft of your operating picture from public signals — what your organization has said publicly about how it works, what it is hiring for, and which systems those roles imply. You get that back at the start rather than being handed a blank questionnaire, and you correct it. Correcting a wrong draft is faster than filling in a blank one, and the corrections themselves tell us where the public picture and the real one diverge.
What gets rated
The loop is assessed across four panels. Each gets its own condition rating and its own confidence level, because a loop can be structurally sound and cognitively blind at the same time, and averaging those two into a single score would hide exactly the thing you are paying to find out.
Panel A
Whether the stated intent still matches the environment, whether the loop is being run as though it were predictable when it is not, and whether options are being held open or quietly foreclosed.
Panel B
How long the cycle from signal to action actually takes, whether context survives each handoff, and whether the gap between what was intended and what the machine inferred is governed by anyone.
Panel C
What coordination costs, who absorbs it, and where a gain booked by one team shows up as a cost on another team’s number. Also where roles have been pulled apart faster than they have been put back together.
Panel D
Whether local performance is being bought at the expense of the whole, whether the loop can repair itself when it breaks, and whether it could survive the load it is forecast to carry.
The ten days
Five phases. The sponsor is involved at three points: agreeing the boundary at the start, the day-eight working session, and the readout. Everyone else we speak to is asked for a single interview. We will give you a specific time estimate when we scope your loop rather than publish an average we have not yet measured.
We agree with you exactly which decision loop is in scope and where it starts and stops. We take in what already exists — procedures, tickets, logs, agent instructions — and map the participants, the tools, and every point where a machine touches the work.
Output: an agreed scope boundary and a first-pass topology
Structured interviews across the loop, deliberately including whoever disagrees. In parallel we reconstruct one or two real decisions end to end, from the originating signal to the outcome, including what got reversed and who overrode what.
Output: decision traces, and the hidden verification costs they expose
We rate each panel, and separately rate how confident we are in that rating. Then we actively look for evidence against our own findings. Anything that survives is reported; anything that does not is recorded as unproven.
Output: panel ratings with confidence stated separately
We walk you through the findings before they are final, and you push back. This is where transferred costs get named out loud — the ones where one team’s improvement is another team’s new problem. It is a working session, not a presentation.
Output: findings pressure-tested against your own knowledge
We assemble the deliverables and present the decision brief. You get what to decide, what constrains the decision, and what the first 90 days should contain — with every finding traceable back to its sources.
Output: the seven deliverables below
How we decide what is true
The failure mode of any short engagement is confident conclusions built on one person’s account. Two rules prevent it, and both are visible to you in the final register.
The sponsor, the people doing the work, whoever sits upstream and downstream, and deliberately the person who thinks the current approach is fine.
Procedures, tickets, logs, the prompts and instructions your agents are actually running, and the record of every exception.
One or two real decisions reconstructed end to end, including reversals, overrides, and the handoffs nobody documented.
The triangulation rule: a finding needs at least two of those three classes to agree. One source is an anecdote, and it gets recorded as unproven and labelled that way rather than quietly dropped.
| Condition | What it means |
|---|---|
| Healthy | Working as intended under current load, with evidence. |
| Strained | Holding, but with costs being absorbed somewhere you can name. |
| Critical | Failing now, or certain to fail under foreseeable load. |
| Indeterminate | The evidence does not support a call. Said plainly rather than guessed. |
Every rating carries a separate confidence level. A critical finding we are only moderately confident about is a different instruction to you than a critical finding we are certain of, and collapsing those two into one number is how diagnostics become sales documents.
What you get
Plain description first, the clinical name underneath. You never need the second one to use the first.
The synthesis, written for the person who has to act on it. Findings, binding constraints, and the decisions that cannot be deferred.
Every human and machine participant, every handoff, where context is stored, and the specific points where it leaks.
The measured latency from signal to action, where the cycle stalls, and whether anything closes the loop back to the people who need to learn from it.
The gains one team booked, set against the costs another team absorbed. This is the finding most likely to change the conversation, because the gain and the cost usually sit on different teams’ numbers.
An explicit authority matrix: what agents may recommend, execute, or block, and exactly where a human must decide. Written down, so it can be argued with.
A sequenced plan with governance attached, scoped so the first moves are achievable without a transformation programme.
The full audit trail. Which evidence supported which finding, what we could not establish, and where you should treat our confidence as low.
Price
Duration does not separate them. Breadth of synthesis does — how many functions the loop crosses, and how many people have to be reconciled to produce one coherent picture.
$4,000
One decision loop inside one function, one accountable sponsor, four to six interviews. All seven deliverables.
$7,500
One loop that crosses functions, eight to twelve interviews, multi-stakeholder synthesis. Adds a cross-team dependency map and a readout built for a board.
Booking is by email for now — we would rather talk to you about the loop before taking your money than have you pick a tier from a checkout page and hope it fits.
What happens after
We tell you which one you got, in writing, with the reasoning. Two of them end with us not working together.
We continue
The findings point at structure: who decides, what the agents are permitted to do, how context moves between teams. That is organizational redesign for metabolic health — how humans and AI agents share work without starving the whole — not a tooling install.
We continue
The findings point at a service-management platform whose licensing no longer justifies itself. Our destination platform is still in development, and we will show you exactly where it stands before you commit to anything.
We refer you out
Sometimes the estate is genuinely worth further investment. We do not do that work, and we will name the partners who do rather than talk you into something that suits us better.
We decline
A short reasons-and-conditions document: why we are saying no, and what would have to be true for the answer to change. The wrong engagement is worse than none, for both of us.