One workflow. Fixed scope. Finished.
Most AI projects stall between the demo and the day real data arrives. Ours are built backwards from that day: what happens when volume grows, when a model call fails, when the person who set it up is on leave. Here is how one workflow goes from a priced loss to a system running live — and what you hold at the end of every stage.
Four stages. Each ends in something you can read, use or count.
01
The method
Four stages, in order, each ending in a deliverable.
You can stop at the end of any stage and still hold something of value.
01
Introductory call
Thirty minutes on the workflow that costs you most. We talk through where enquiries enter, where they stall, and what a diagnostic would need to look at. Nothing is committed on the call.
You leave with a clear picture of what the diagnostic would cover, what it would cost, and what you would hold at the end of it.
02
Paid diagnostic
A few days inside the real inbox and the real workflow. We read the messages, watch the handoffs, and count what falls through. The deliverable is written findings with the loss priced in your own numbers.
Yours to keep and act on, whoever you build with. If the loss turns out smaller than you feared, the findings say that too.
03
Fixed build
One workflow, end to end, at a scope and price fixed in writing before work starts. Weeks, not quarters. Instrumented from day one so the effect is visible rather than asserted.
Response time, conversion and hours returned are measured from the moment it goes live.
04
Measured results
The instrumentation is read against the baseline the diagnostic set: how long an enquiry waits now, hours returned to the team, what converts. Before and after, in the same units.
The effect shows up in your own numbers — the same ones the loss was priced in.
02
What we commit to
Fixed before you spend anything.
Each of these is in the paperwork before work begins — not negotiated afterwards.
30 min
The introductory call. No preparation needed.
Fixed
Scope and price agreed in writing before work starts.
1
One workflow per build. End to end, not a pilot.
Weeks
From agreed scope to a system running live.
Written
Findings you keep, whoever builds it.
Yours
Code and data ownership, in writing.
03
Principles
The rules the work is built on.
Each one exists because it protects the result you are paying for.
Diagnosis before a quote
You should not have to guess what to buy. We read the workflow first, put a figure on the loss, and only then propose a build — so the scope rests on evidence, not on a hunch.
Priced in your own numbers
The loss is measured from your records, in your currency, with the assumptions written down. That figure is the line any build is judged against.
One workflow at a time
Narrow scope finishes. We take the single path where money leaves fastest, build it end to end, and get it live before anything else is discussed.
Built for the bad day
Retries, fallbacks, rate limits, logging and a human handover path are part of the build, not a later phase. The system is designed for the day something fails.
Measured, not asserted
Response time, conversion and hours returned are wired in before launch, so you can see what changed without taking our word for it.
Handover, not dependency
Running code, credentials, a runbook and a walkthrough. You can keep the system without us — staying is a choice, never a trap.
04
What is not in the deal
Things you will never be sold here.
Knowing what is not in the engagement is half of trusting it.
No software licences. You own the code outright on final payment, and you pay tool vendors directly, with no mark-up.
No open-ended retainers as a condition of starting. Support after handover is a separate agreement you can take or leave.
No dashboards nobody opens. Measurement is limited to the figures the build is actually judged against.
No strategy documents. The diagnostic ends in findings you can act on that week, not a vision deck.
No pilots that never leave the lab. The build is one production system, live in your business, or it does not ship.
No lock-in. The handover includes the code, the credentials, the runbook and a walkthrough — everything needed to run it without us.
Next
Start with thirty minutes.
Bring the workflow that annoys you most and a rough sense of what it costs. By the end of the call you will know whether there is a loss worth pricing, and exactly what finding out would involve.
