AI Product Build
We decide what to build with you, then design, build and launch it. Fixed scope, fixed end date, and your code in your repository from day one.
Something real to build, and one chance to build it right.
You're a funded founder or a domain operator with a product to launch. Or your team has a prototype, often made with AI coding tools, that won't survive real users.
Strategy, UX, build and launch come from one team, so the decision about what to build and the build itself never drift apart. That includes turning an AI-built prototype into a production product.
A product that works on real data, and the reasoning behind it.
A success metric
Defined before any architecture.
The model-versus-rules decision
Which parts need a model, and which work better as rules.
A working prototype on real data
With an evaluation set and a go or no-go recommendation.
The production application
To the agreed scope: backend, data layer, retrieval, agent orchestration, integrations, front end, auth and admin tooling.
Testing that includes the model
Regression suites before every deploy, and adversarial tests for prompt injection and data exfiltration.
A graduated launch
With tracing, cost and quality dashboards, and launch assets.
Documentation and a handover
With the reasoning behind each choice written down.
Seven stages, each with a gate.
Nothing moves to the next stage until the gate is met. A failed prototype is a cheap discovery, not a failed engagement.
Built with this service
Our own ventures, built and run by us. Client product work appears here with permission.
Fixed fee, milestone-based.
Fixed fee, set after scoping, in writing before work starts. Payments are milestone-based, as set out in your proposal. You can stop at any milestone and pay only for the milestones completed.
Product Care is optional after launch: monitoring, fixes, and one small improvement a month.
Four things, up front.
Data access
The prototype runs on real data, not samples.
A technical counterpart on your side
Decisions within days, not committees
Constraints stated up front
Compliance, security, integration, latency and cost ceilings.
Not a demo in production clothes.
Not this: A demo.
Instead: A production application, built to the agreed scope and working on real data.
Not this: A proof of concept dressed as a product.
Instead: A prototype tested on real data first, with an evaluation set and a go or no-go recommendation.
Not this: A build shipped without evaluation or observability.
Instead: Regression suites before every deploy, and tracing, cost and quality dashboards at launch.
We have a partial build, or a prototype made with AI tools. Can you work with that?
Usually, yes. Much of our work starts with something already built. We assess what exists before committing to extend it. Sometimes the right call is to build on it; sometimes the architecture blocks where you're trying to go, and restarting is cheaper. We'll tell you which, with the reasoning, and you decide.
Who owns the code and the work?
You do. Code lives in your repository from day one, and IP transfers to you on payment.
What if it doesn't work out?
You can stop at any milestone, and you pay only for the milestones completed. We'd rather find out early than late.
Can we negotiate the price?
Scope is negotiable; the price follows the scope. If the number doesn't work, we'll show you what comes out to reach it. If what's left won't produce the outcome, we'll tell you that instead of quietly shipping a thinner version.
Bring the product. We'll tell you what to build first.
We reply the same business day.