AOS for Product and Strategy Decisions

We have pointed this Council at security and at architecture. Engineering decisions, judged on engineering evidence: a threat model, a blast radius, a sub-millisecond session read with headroom to spare. Now the harder proof. The same profile, the same eleven specialists, the same neutral Arbiter, aimed at a pricing decision, where the evidence is willingness to pay and churn risk, not test coverage. If the machinery only works on engineering questions, it is not general. It is a niche tool with good marketing.

I ran the `strategic-council` profile against a seat-to-usage pricing decision, on the `saas` domain. Session 2026-07-01-strategic-council-mr2bkvpf, five rounds of debate, and this time the memo rendered clean to disk. Not one DORA metric in it.

Where the evidence standard lives

Issue 4 made a claim that pays off right here: evidence standards do not live in the profile, they live on the agents. Catalyst's agent.yaml says it is convinced_by revenue data and willingness-to-pay signals. That line did not change between the auto-deploy run in Issue 3 and this one. What changed is the question you point it at.

Ask Catalyst about autonomous deploys and its speed-and-monetization bias argues deployment frequency against change-failure rate. Ask it about pricing and the exact same bias argues revenue velocity against the competitor's conversion clock. Same bar, different currency. Sentinel still guards durability and trust. Here that means something specific: a customer who never opens a bill they did not see coming. Navigator reads positioning off a competitive window instead of a threat surface. And Advocate still speaks for the person on the other end, except that person is a buyer with a budget, not a user with a session.

The frame around them never moved. The Arbiter runs the room. [catalyst, sentinel] is still the load-bearing tension pair. Here that tension is upside against churn; in the security review it was velocity against blast radius. The Provocateur still speaks last, and the documented dissent still goes on the record. I pointed the same eleven agents at a pricing call and never touched a line of their config.

The four who earn their keep here

A pricing call is where Catalyst, Navigator, Sentinel, and Advocate stop being extras and become the show. Each one showed up with a bias built for exactly this.

Catalyst brought the clock. The company is losing heavy accounts that underpay and small accounts a pure-usage competitor is converting in real time.

We're bleeding on both ends now. The "safe" seat-surcharge hands the competitor their own price comparison.
Catalyst

Navigator refused to treat pricing as arithmetic.

Whoever owns "here is what you pay for" owns the category. The floor is a forecastability weapon the pure-usage competitor structurally cannot copy. Sell it to the customer's CFO.
Navigator

Sentinel found the trap under everything: the accounts you want to charge more are the accounts you cannot afford to lose.

The under-charged accounts are the reference logos. The revenue case and churn case point at the same population. Treat their transition as a trust operation.
Sentinel

And Advocate spoke for the buyer against the instinct to meter every tick.

Buyers renew what they can forecast, not what feels "fair." Per-action metering triggers self-throttling that strangles the heavy-workflow value loop we're monetizing. Trust behavior, not survey quotes.
Advocate

Four biases, four kinds of evidence, none of them a metric an SRE would recognize.

The decision

The brief fills the profile's four sections (## Situation, ## Stakes, ## Constraints, ## Key Question), the same shape as every strategic-council run. A B2B SaaS on seat pricing. Its biggest accounts buy the minimum seats and share logins, because their value is in a few automated workflows, not headcount. A competitor prices purely on usage and is peeling off the small, high-activity accounts. Series B board, net revenue retention has to climb, and it cannot absorb a churn spike this year. The question: move to usage pricing, and if so, how do you sequence it to capture the underpriced heavies and blunt the competitor without churning the logos you sell against?

Real tension, and it maps straight onto the profile's pairs. Catalyst against Sentinel is upside against churn. Advocate against Navigator is how customers actually buy against how the market gets positioned. This is the room those pairs were built for.

The debate

Ten of the eleven speaking agents agreed almost immediately, and the Arbiter treated the speed of that agreement as a warning.

Within four minutes, ten of eleven speaking agents had converged on the same shape. Unanimity that fast is a diagnostic signal of shallow consensus, not depth.
The Arbiter, deliberation summary

Two agents had already planted charges under the easy answer. Pathfinder called the obvious "seat plus a usage surcharge" hybrid an incremental trap that anchors your price on a dying unit, and argued for billing the completed outcome instead of the raw run.

Bill on runs and we validate the competitor's wedge at a higher price. Bill on outcomes and they become the commodity meter that charges for failures.
Pathfinder

The Provocateur, speaking last, went after the board promise everyone was leaning on. A committed floor makes revenue look smooth. It does not.

The floor smooths the base, not the marginal dollar, and the marginal dollar is the whole point. Higher-mean and higher-variance. Do not sell one hiding the other.
Provocateur

Between them, Pathfinder and the Provocateur are why the fast consensus did not hold. The most productive tension, by the Arbiter's own read, was Catalyst against Provocateur: speed against scrutiny, and neither won. They resolved into a reframe. The fastest way to stop the bleeding and the most cautious thing you can do turned out to be the same move.

The memo, with the dissent kept

Here is what came back, and it is not "change your pricing."

The decision on the table this quarter is not "change our pricing." It is run a two-week, read-only, falsifiable test whose result we pre-commit to obey.
The Arbiter, the one-sentence answer

Recommendation #1 is that read-only shadow-bill, run over meter data the company already holds. No money moves, no customer sees anything. It recomputes, on the same historical period, the invoice each account would have received under three units: per-run, per-completion, and seat-equivalent. Pre-register the pass and fail thresholds before week one, and name the person allowed to call it off, because a gate that can only say yes is a launch with a waiting room. Strategist gave it the line the Arbiter said it would stake its name on: commit to the test, not the transition.

A decision that is cheap to reverse is the only honest one we can hand them.
The memo, on Recommendation #1

Being wrong costs two weeks and zero customers instead of a quarter and a reference logo, and this is a board that cannot survive a churn spike.

Recommendation #2 is the destination the test is allowed to disprove: kill the phrase "seat plus surcharge," denominate the floor in committed dollars, meter overage on a completed outcome, and structure the whole thing as an annual take-or-pay commitment with drawdown and rollover, the committed-capacity pattern Snowflake sells on. The outcome unit is also the competitive reframe. We bill only when it works. They bill you for every retry.

Then the part a solo strategy call would have smoothed away. The dissent, kept on the record, in ink.

The floor-to-total ratio, the single number that decides whether you have rebuilt seats with extra ceremony or handed the board a casino, was never set. Provocateur flagged it as the open wound and nobody closed it. Advocate warned that heavy accounts will feel every tick of the meter and throttle themselves, so the committed floor has to cover their dependable volume and only the increment can be metered. And Sentinel drew the bright line the plan lives or dies on.

Protect NRR with structure, never a punitive floor the customer knows they can't consume.
Sentinel

The test the memo hands the board for that line is one question: would the customer choose this if they read it carefully? Auditor surfaced one more: the company has no retrieved record of its own last pricing change, so it is making its largest pricing bet with a blank memory of the previous one. And Operator reduced everything to one binary. Can you name a billing-eng owner and a RevOps owner today, with cleared calendars and freeze authority? If not, it is not a plan yet, just a direction, and you say so to the board.

None of that resolved, and the memo keeps it on the record rather than smoothing it away.

Why this matters for the whole series

Issue 1 made a claim I have been paying off ever since: the harness makes Councils reusable across domains. A profile that only argues engineering questions is a niche tool. One that runs a pricing decision on the same Arbiter and tension pairs, and keeps the dissent on the record exactly as a security review does, is a general instrument.

That is what this run proves, in the domain furthest from where the harness started. This run is all willingness to pay, churn, competitive timing, and how real buyers behave, held to the same standard of proof and argued by the same machinery that argued autonomous deploys eight issues ago.

The Council does not know it left engineering. The evidence standard changed, the machinery never noticed.

Next

Eleven issues watching profiles the harness ships. Next issue you build your own. We author a profile from scratch, for a domain nobody has covered yet, and push it to the repo. The docs walk through it. Issue 12, the graduation.