AI should solve a real problem, or wait
We build custom AI tools around your own data when there is a defined problem and a measurable result, and say so plainly when there is not.
AI is changing how buyers find suppliers, what they expect to ask a website, and what a small firm can build for the price of a conventional project. A business that treats that as a passing feature will be caught out; one that treats it as magic will be sold something that does not work. We do two things about it.
We build custom AI tools around your own data
Engineering projects with a defined problem and a measurable result. The kinds of problem that qualify:
- A question your customers ask that your data could answer, if it were in order. Which of our 400 colours matches this sample; which product suits this substrate at this temperature.
- A task your team does by hand that follows rules buried in documents. Turning a specification into a quotation; checking an order against a data sheet.
- A body of material nobody can search properly. Twenty years of technical notes or test results.
Each starts with the problem in a sentence, the data it needs, and the number that will show whether it worked. The tool runs in your account, on your data, and we show you what it got wrong as well as right.
What we promise, and you can check
- Every project starts with a defined problem and a number that will show whether it worked.
- If it won't pay off yet, we say so, in writing, and what would have to be true first.
- Your data stays yours, and the tool is in your account.
We say when it won't pay
Often the answer is not yet: the data is not in order, so a tool would be confidently wrong; the problem is rare enough that a person is cheaper; or what the business needs is a product finder, which is not AI. Industrial customers, in the one survey that asks them, rate product selectors most useful and chatbots least, and their trust in generative AI is below five out of ten. The advice is written, names what would have to be true first, and costs a conversation.
What we won't do
No chatbot on your homepage. No project whose success cannot be measured. Nothing described as powered by AI, because what matters is whether it answered the question.
How we price it
A scoping conversation costs nothing. A project is quoted after it, with the problem, the data, the measure and the price on one page.
Last reviewed 28 October 2026.
Tell us the problem in a sentence.
An engineer will reply within one working day with whether AI is the right tool for it, and if not, what is.
Talk to an engineer about the problem