Nobody should buy AI on the strength of a demo
Every demo works. Demos are built on clean data, on a happy path, by the person who wrote them. Your business is none of those things.
- It is small on purpose. One process, one measurable outcome. A wrong answer costs you weeks, not a year’s budget.
- It runs on your data. Your documents, your exceptions, your volumes. That is where AI projects fail, so that is where we start.
- It ends with a decision. A measured before-and-after and a written recommendation — including, when it is right, not to proceed.
Five steps, one fixed period
Dates and scope are agreed in writing before we start. Nothing expands halfway through.
- 01ApplyTell us the process and what it costs today. Ten minutes.
- 02Scoping workshopOne session with the people who own it. We agree the measure.
- 03BuildA working solution on your data, with weekly check-ins.
- 04Run and measureYour team uses it for real, beside the current process.
- 05DecideWritten result and recommendation. Scale, park, or stop.
↑ The step a pilot without a decision never reaches
What we ask, and what you get back
Printed as a pair because they are one bargain. Neither column works without the other.
Four things from you
- A named owner for the process who can decide how it should work. Not just an IT contact.
- Access to real data, under whatever protection agreement you need. Synthetic data proves nothing.
- A few hours a week from the people who do the work.
- Honest feedback, especially when what we built is annoying to use.
Four things back
- A working solution on your own data. Not a prototype switched off at the end.
- A measured before-and-after against week one, in a form you can put to a board.
- A written recommendation: what to do next, what it costs, and what we would not do.
- Everything we built, documented, whether or not you continue with us.
What is not settled on this page yet
To confirm before launch. The commercial terms of the test bed — duration, price or whether places are free, how many places are available, and the eligibility criteria — still need to be settled and stated plainly here.
Prospects will look for them immediately, and vagueness costs more applications than an unwelcome number would.
We measure the process as it runs today, before we change anything

Tell us about the process
No obligation, no mailing list. If a test bed is not right for you, we will say so and explain why.
Sent
Thank you â your application has arrived
Every application gets an answer, including the ones we turn down, and the answer says why. If what you have described needs more detail before we can judge it, we will ask rather than guess.
Before you start
About ten minutes. The more specific the process, the more useful our reply.
If you would rather talk it through first, that is fine too — the call comes to the same place.
Talk to us instead