You own everything we build.
It runs in your cloud account. If you stop paying us, it keeps running. What stops is us watching it, fixing it and extending it. We do not publish a price, because we do not quote a build we have not scoped. The audit is a fixed fee. The build is a fixed fee, quoted only after the audit. The run is monthly.
You can stop after the audit and take the spec elsewhere
The runbook is yours the day it is delivered. It names the systems, the triggers, the exceptions and the decisions your people actually make, in enough detail that a different team could build from it.
You own what we build. Nothing on this page is a subscription to software. The monthly fee is for watching, fixing and extending the system, not for permission to use it.
Send us one thing your team repeats.
Nothing to prepare, and no documents to send first. The meeting is forty five minutes, and you leave with a one-page sketch of the job you brought. If it is not worth building we say so.
The audit
The audit comes in two depths.
The gates, the test set and the cost model are audit deliverables. That is what makes the build quotable instead of guessed.
The Map
The operation mapped with its exceptions, the jobs ranked worth programming and not, the value case in your own numbers, and an honest no where AI is the wrong tool. Yours to keep either way.
The Spec
Everything in the Map, then the operation compiled: the workflow written step by step, your thresholds and approvals as executable rules, the line the system will not cross alone, the test set built from your own hardest cases, and the running cost modelled per step before anything is built. A build another team could execute.
Every job tied to revenue or cost, in your own units, signed off by the person who does the job.
During the audit the function is baselined in your own units: volumes, cycle times, exception rates, hours by step from observation, win rates. Then exactly two lines. Revenue events: quotes answered in the hour, bids submitted, renewals caught, claims recovered. Cost lines: hours at loaded cost, error cost, leakage. Ranges come from your own data, signed off by the person who does the job, and expressed per shipment, per bed, per deal. Never percentages.
Questionnaires cannot produce this. Weeks inside the function can.
Three smaller ways in
Not every engagement starts with the audit.
Your team, enabled
We install Claude or ChatGPT across your team, set the permissions and the governance, and teach the habits that make it stick. Then you watch which jobs your people hand it first. That list is the start of your map.
The AI you already built, proven
You built a copilot, or paid someone to. Nobody can say if it is safe to trust. We test it against your own hardest cases, name every way it fails, and hand you the test set that decides when it ships. If it should be switched off, we say so.
The AI bill, brought down
Your AI works and the bill grows faster than the usage. We baseline what every workflow spends, route each step to the cheapest model that passes your quality bar, and prove nothing got worse. The savings are measured in your own invoices.
Each is a fixed fee, smaller than the audit, agreed in writing before it starts.
