Eleven weeks from first call to a system acting on its own.
Three weeks of diagnosis, eight weeks of build and proof, then an operated system whose autonomy you raise a rung at a time. Nothing here is a phase we invented to fill a proposal - each step exists because skipping it produces a specific, expensive failure.
Three phases, and a decision point you can stop at.
Diagnose, then build and prove, then run. The audit ends at a decision, not at a renewal - and one of the three ways it can end is that we tell you not to build anything yet.
Eleven weeks, and two of them are exits.
Diagnose, build and prove, then run. Each phase says what it costs you, what you leave with, and whether you can stop there. Two of the three ends with you keeping everything and owing nothing further.
1DiagnoseThree weeks
Discovery
Read-only access to your CRM and revenue tooling. We map what your revenue engine actually does, which is rarely what the process document says it does.
Analysis
45 metrics scored and benchmarked against companies your size, with every gap you have priced in dollars rather than described in adjectives.
The report and the map
Walked through live in 90 minutes, with whoever will own the fix in the room. A PDF in an inbox is not a finding, it is a file.
Two hours of one person's time, and read-only access.
The diagnostic, the benchmarks and the Operating Map.
Here. Keep all three, owe nothing further.
2Build and proveAbout eight weeks
Scope and golden dataset
We write down what the system may and may not do. You pick and label the twenty historical cases it will be graded against, because if we chose the answers the test would only measure our opinion of our own work.
Build
Inside your stack, not beside it. Every decision is tagged RULES, AI or HUMAN, and anything a rule can do, a rule does: cheaper to run, faster, and possible to reason about at 2am.
Prove
Graded against your twenty cases and written up in the Test Report. Below 85%, or one uncaught unsafe action, and it does not ship. We fix and re-run rather than launch and watch.
Go live at rung one
Live data, no actions taken. It logs the decision it would have made beside what actually happened, and you read the difference for a fortnight before it is allowed to touch anything.
Twenty labelled cases, and a named owner.
The system, the Test Report and the evidence log.
After the report. It is yours either way.
3RunFor as long as it earns it
We operate what we built
Priced by systems under management, not by hours, so a system that needs less of our time costs you less rather than quietly billing more.
Autonomy is raised, never assumed
Watch, then approval, then conditional action. Each rung is earned against logged evidence, and every one of them is reversible.
Three accounts at a time
If that sounds like a small number, that is the point. It is what makes the operating part real rather than a retainer with a dashboard attached.
A monthly fee, and a named owner on your side.
An operated system and a rising autonomy rung.
Any month. The system stays with you.
Three ways an audit ends. We earn nothing from one of them.
One recommended System Build
The most common ending. One named system, fixed price, about eight weeks, proved against your own history before it touches anything live.
A Data Readiness Sprint
Cleanup, enrichment and routing, with no AI in it, and we say so plainly. Putting a model on top of bad data does not fix the data, it industrialises the error.
Do not spend money yet
You keep the map, the benchmarks and the baselines, and we build nothing. We would rather lose the work than sell a system that fails its own test.
A firm that names the ending where it earns nothing is a firm whose other claims are worth checking.
One control stops the system, it is yours, and it does not require us.
An emergency stop that goes through a support queue is not an emergency stop. It is a support queue.
What you own, whatever you decide.
The Operating Map is yours from week three
Whether or not you build anything with us, and whether or not you build it with someone else. It is a deliverable, not a hostage.
The system is yours at the end of the build
It runs in your stack, on your accounts, under your credentials. There is nothing to migrate if you stop working with us.
Nothing you give us trains anything
Your data is used to build and grade your system. It does not become a feature for the next client, and it does not train a model.
You can leave
Any month, without a renewal window or an exit project. The exit is boring by design.
It starts with three weeks and one number
Which leak is costing you most. Everything on this page is a decision you make after that answer exists, not before.