Engagement 02
One use case off the catalog, shipped to production. Deployed on your infrastructure, monitored, documented, and handed to the team who will own it. Fixed price, agreed before a line of code is written.
The shape of it
A Build starts where the Scan finished: a use case that has already been scored, estimated, and demonstrated as working software. That is the whole reason the price can be fixed and the timeline can be short — the unknowns were spent in the Scan, at a fraction of the cost.
One senior lead owns the engagement and stays the person you hold to the date. The work is delivered by a small team of senior engineers we have worked with before and pick for the specific problem. There is no bench to keep busy and nobody learning on your budget.
Production means production: running under your identity provider, inside your compliance boundary, monitored, and documented well enough that your engineers can change it without calling us.
Ten weeks shown. Most Builds are shorter; the Scan estimate says which.
What gets built
Almost everything worth building in a mid-market company falls into one of these. The catalog from the Scan tells you which one your best candidate is.
A multi-step internal process with handoffs, waiting time and rekeying between systems — collapsed into one flow with a human review at the point where judgement is actually required.
For example — Client onboarding packs assembled from source systems and checked once, instead of assembled by hand and checked four times.
An agent doing a defined job continuously, inside your boundaries, with logging and an escalation path. Not a chatbot bolted onto a website — a worker with a scope, a supervisor and an audit trail.
For example — First-pass review of inbound documents: extract, check against policy, flag exceptions to a named person, file the rest.
A per-seat tool replaced with an application that lives in your infrastructure and does the specific 20% your people actually use. Running cost becomes tokens and hosting rather than a licence that grows with headcount.
For example — A seat-priced reporting or workflow tool retired in favour of an owned app that produces the same output from the same data.
Fixed price
A fixed price is only honest if both halves of this are written down before you sign. Here is our half, in advance.
Third-party costs
Model API usage, cloud hosting and any licences run on your accounts, at cost. We do not resell them and we do not mark them up.
Work outside the written scope
If something genuinely new surfaces mid-build, we price it as a change and you decide. The alternative — absorbing it quietly — is how fixed prices become fiction.
Data remediation beyond the estimate
The Scan establishes whether your data is ready. If it is materially worse than it looked, that is a scoped piece of work in its own right.
Running it after handover
Handover means your team can operate it. If you would rather we kept owning it, that is Run.
The risk that sits with us: if the work turns out harder than we estimated, the price does not move. That is what you are buying, and it is why we will not quote one without a Scan behind it.
Price
$30,000 – $80,000
Fixed per engagement, scoped from the Scan. 6–10 weeks.
The range is honest rather than decorative. A single-system automation over clean data sits near the bottom of it. A workflow spanning four systems inside a regulated process, with an audit trail somebody will inspect, sits near the top.
What moves the number is scope, not our availability: how many systems the work touches, how clean the data is, and how hard the compliance boundary is. All three are established during the Scan, which is why the quote you get afterwards is a number rather than a range.
Then judge it against the alternative. One senior engineer, fully loaded, costs more than the top of this range in their first year — and cannot start for three months, will not have shipped anything by month six, and is still there if the use case turns out to be wrong.
Questions
Three things, in order: how many systems the work has to touch, how clean the underlying data is, and how hard the compliance boundary is. A single-system automation over tidy data sits at the bottom of the range. Something spanning four systems inside a regulated process sits at the top. The Scan is what turns that from a guess into a number, which is why we quote a Build from a Scan rather than from a conversation.
Sometimes. If you already know the use case, the data is understood, and the acceptance criteria are clear, then a Scan would mostly tell you things you know. What we will not do is quote a fixed price on a problem neither side has examined — that price would have our uncertainty priced into it, and you would be paying for it.
You own it outright: the code, the infrastructure definitions, the documentation, all of it, in your repositories from day one rather than transferred at the end. There is no runtime dependency on us, no licence back to us, and nothing in the stack that only we can operate. Handover is a deliverable with acceptance criteria, not a goodwill gesture.
Because of how the work is produced. AI agents draft the requirements, the code and the tests; senior engineers judge, harden and ship them. The scarce input is senior judgement rather than typing hours, so adding mid-level headcount would not make it faster — it would make it slower and more expensive. It is also the method the fixed price is underwritten by: if it does not hold, the overrun is ours.
Fair question to ask of a small team. Your code and infrastructure live in your accounts from day one, the documentation is written as we go rather than at the end, and every engineer on a Build is senior enough to carry the work forward alone. The dependency you are taking on is on the artefacts, not on any one person’s availability.
Next step
If you already know the use case and the data is understood, say so on the call and we can talk about scoping a Build directly. If you do not, the Scan is the faster route to a price both sides can trust.
No deck, no discovery questionnaire, no sales engineer on the line. You talk to the people who would do the work.