The Seat Audit
Two passes down one licence bill — who the seats are actually for, and what they cost once nobody is subsidising them.
Thirty minutes, one bill. Pass one asks whether any named work runs differently. Pass two asks whether you would buy it at three times the price. Every line lands in theatre, countdown or load-bearing, and the bill takes the name of whichever column holds it.
Two passes, in order — the second is meaningless without the first
Pass one — does the work run differently?
- The named-work test. Name a piece of work that runs differently because these people hold this licence. No name — the line is theatre, and pass two never applies.
- The capability test. Do the people holding these seats have both the work and the skill to use it? Provisioned-because-it-was-easier is not a population.
- The receipt test. Could somebody state the change in units — cycle time, cost, error rate? "Feels faster" is not a receipt.
Access is not change. A seat buys somebody the option to use a model. It does not remove a step, delete a handoff or cut a cycle time. That is why utilisation charts stay green while the P&L does not move.
Pass two — would you buy it at full price?
- The triple test. At three times today's price, would you renew this line? Three is not a forecast — it is the shape of a bill with nobody else paying part of it.
- The seat-count test. Does it pass for everybody on the line, or for a handful? Passing for a handful means the seat count is wrong, not the tool.
- The denominator test. Can you state this line's cost per unit of work — per claim, per quote, per job? No denominator, no defence at renewal.
Your price is somebody else's loss. OpenAI earned $13.07 billion in 2025 and ran a $20.92 billion operating loss — roughly $2.60 spent per dollar earned. Your renewal price depends on how long investors keep funding that gap, which is not a variable you control.
The three landings
- Theatre. Fails pass one. Cancelling changes nothing except the bill.
- Countdown. Passes one, fails two. Real work runs differently — but only while the price is subsidised. There is a clock on it you did not start and cannot see. The dangerous column: it looks like success on every dashboard you own.
- Load-bearing. Passes both. Worth buying at three times the price. The business leans on it and it does not move.
Cancelling is not the strategy. The money freed here becomes a build bill: a one-off, token-hungry spend using agentic tooling to construct the process around its judgment points — including the processes that were never automatable before, because one read-and-decide step in the middle blocked the whole chain. What is left afterwards is a run bill: deterministic steps at effectively no cost, calling a model only at the judgment points. Spend generously once; pay a fraction of a cent per transaction after.
The bill
| Licence line | Seats | Pass 1 · work runs differently? | Pass 2 · buy it at 3×? | Landing |
|---|---|---|---|---|
| — | ||||
| — | ||||
| — | ||||
| — | ||||
| — |
Verdict
—
Classify at least one line with seats on it.
The bill takes the name of whichever column holds it. Theatre holds most seats · theatre — you are paying for access, not outcomes. Load-bearing holds half or more · load-bearing — the bill earns its keep at any plausible price. Anything in between · countdown — real work, on a clock somebody else set.
One boundary, honestly: this audit cannot tell you what tokens will cost next year. Nobody can, and no vendor will volunteer it. What it removes is the assumption — which lines already earn at any price, which live on somebody else's losses, and which were never doing anything at all.
The Seat Audit · Peak State Global · peakstate.global — from the essay on seats and decisions. Print this page for the worksheet version; the verdict box travels with it. Its companion is the Judgment-Point Map.
Bring us the bill as it stands
Landed on theatre or countdown? That is the useful outcome — it means the money is already there, pointed at access instead of outcomes. A 30-minute conversation about the first judgment point worth funding is where the conversion starts.
The Seat Audit is one instrument in the Peak State field kit — essays on the leader's job as it is shifting, and the tools that go with them.
All essays and instruments →