The scope is written down
If it is not in the scope document, it is not in the build. Changes are welcome, priced and re-signed — never silently absorbed.
Home / Process
How we run
Five phases,
nothing hidden.
A typical build runs eight to ten weeks from first call to production. Here is exactly what happens in each week, and what you hold at the end of it.
Scope
We interrogate the problem before the solution. What breaks today, who feels it, what a fix is worth. You leave this phase with a written scope and a number, not a vibe.
You leave with
- Written scope
- Success metrics
- Fixed-price estimate
Architect
Data model, integrations, failure modes and the boring decisions that are expensive to reverse. For AI work, the eval set is built here.
You leave with
- System diagram
- Data model
- Eval harness
Build
Weekly demos against a staging environment you can click. No dark period, no big reveal. If a decision needs you, you hear about it that week.
You leave with
- Staging environment
- Weekly demo
- Test suite
Ship
Production deploy, monitoring, alerting and a runbook written for your team. Then we watch the graphs with you for the first fortnight.
You leave with
- Production deploy
- Monitoring & alerts
- Runbook
Operate
Optional. We keep the thing running — patches, model swaps, cost tuning and the next set of features, on a monthly retainer.
You leave with
- SLA & on-call
- Monthly report
- Roadmap reviews
Ground rules
Four things we do not bend on.
You see it every week
A clickable staging environment and a short demo, every week, from week three. No status reports about work you cannot inspect.
The eval comes first
For anything AI-shaped, we build the measurement before the feature. Otherwise nobody can tell whether it is getting better.
You own everything
Your repositories, your cloud accounts, your API keys. We work inside your infrastructure and leave a runbook behind.