How we work
Problem. Prototype. Approve. Build. Scale.
Five steps, in that order, every time. The prototype comes before the commitment, which is the whole point. Here is what each step does, what it needs from you, and where it stops.
01
Problem
Tell us what is slow, manual or missing
Say it the way you would say it to a colleague. A workflow, a pain, an outcome you want. Nobody sends you a requirements template.
02
Prototype
See something you can click, within hours
Two to three hours after the conversation, a working prototype you can test, break and argue with. Not a document, not a deck.
03
Approve
Decide with the thing in your hands
You give feedback, we change it, you decide. If the prototype shows the idea was wrong, that is a cheap and useful result.
04
Build
The approved prototype becomes the real application
Engineering, integration with what you already run, testing with your people, and a go-live. Fixed scope, agreed before we start.
05
Scale
Extend it, and measure what it did to the people
More users, more workflows, more intelligence. And a number, next to the hours saved, for what your people did with the capacity it created.
The Coffee Test.
You describe the problem over one coffee. Two to three hours later, there is something on a screen for you to react to. Not a document. Not a deck. Something you can click. If there isn't, we haven't understood the problem yet, and that is on us.
From approved prototype to production
Seven days. Scoped on day zero.
The prototype arrives within hours. Once you approve it, the build runs to a fixed weekly rhythm. Here is what each day is for.
- Day 0 You bring the requirement. Two to three hours later, the prototype is in your hands.
- Day 1 You break it. We listen. You approve.
- Day 2 Engineering begins on the approved prototype.
- Day 3 Engineering.
- Day 4 Integration with what you already run.
- Day 5 Testing, with your people.
- Day 6 Fixes and sign-off.
- Day 7 Production-ready.
When the problem is an AI problem
Forecasts and recommendations need one extra step: the data has to be able to answer the question.
For an application, the prototype is the proof. For a prediction, the proof is your own history. So the intelligence track adds a readiness check before anyone builds anything.
-
01
Assessment call
30 minutes · free
You describe the process. We ask the questions that establish whether the data can support a useful answer, and you leave with our honest view of where the highest-value starting point is.
-
02
Data readiness check
A few days
We look at a sample of the relevant data and tell you what it can and cannot support. Sometimes this is where we stop, because the history needed to answer the question does not exist yet.
-
03
Prototype, then scoped build
Days, not quarters
A prototype on sample data within two to three hours, free, so you see the shape of the thing. Then a fixed-scope, fixed-price build against one written success measure, reviewed with you as it goes rather than revealed at the end.
-
04
The honest verdict
Agreed in advance
We measure against the number we agreed, and we tell you what it says even when the answer is that this was not worth doing. That finding has value too, and you get it either way.
Is this for you
A good fit, and an honest not-yet.
We would rather say this here than three meetings in.
Probably a good fit if
- You can describe one workflow, pain or outcome, even without a precise figure
- Someone senior can approve a small, contained piece of work
- You would rather react to a prototype than write a specification
- You are comfortable being an early customer of a young company
Probably not yet if
- Procurement requires three reference customers and a certification pack
- The goal is a company-wide AI strategy rather than a specific problem
- The decision you want automated is clinical, diagnostic or legal
- Nobody internally can spare a few hours during the build