How we work
You should be able to fire us cheaply.
Every stage below is small enough to walk away from. That is deliberate — it is the only honest way to sell work whose value you cannot verify until it exists.
- Week 0
Introduction
Email us about your business and the problem. You get an honest first opinion on whether it is worth solving with technology — including when the answer is no.
Produces: A one-page written assessment
- Weeks 1–2
Scoped build
One workflow, fixed fee, acceptance criteria agreed before we start. Not a slide deck and not a pilot that dies in a sandbox.
Produces: A running system in your stack
- Week 3+
Run it or take it
We hand it over documented and you own it, or we keep it running on a light monthly retainer. Decided at the end, not locked in at signature.
Produces: Docs, runbook, and your choice
What we need from you
Less than you would expect, but it is not nothing. A build goes badly when these are missing, so we would rather say them up front:
- One person who owns the process. Not a committee — someone who can say how it actually works today and decide when it is right.
- Access to real data. Sanitised is fine, synthetic is not. Automation fails on the messy cases, and synthetic data never has them.
- An honest answer on volume. How often the process runs, and how much it costs you when it goes wrong. That is the whole business case.
What we will tell you no about
A short list, because it saves everyone a meeting:
- Automating something nobody has written down. If the process only exists in one person's head, the first job is documenting it — and you do not need us for that.
- Anything where a wrong answer is unrecoverable and no human approval step is acceptable. We will not put a model in that position.
- Work that is a rules engine wearing an AI costume. We will say so, and point you at the cheaper thing.