Insights

How we think about the work

No vendor talking points and no thought-leadership filler. Practical writing on the technical decisions we have spent our careers making — including the ones we got wrong first.

Latest AI & Machine Learning 6 min read

The golden dataset: why an AI build should start with 300 labelled examples

Before any model is built, an AI project needs a few hundred real examples with agreed correct answers. The number matters, the sampling matters more, and writing the answers turns out to be writing the specification.

Read the article

Integrating with systems that have no API

“No API” rarely means no way in. It means no documented one. What matters is how far from the screen you can get, and whether you are reading from the system or writing to it.

Replacing the spreadsheet only one person understands

The formulas are the easy part. What you are actually replacing is a specification that exists in one person’s head, and the projects that fail are the ones that never went looking for it.

Why most AI pilots never reach production

The demo is the easy part. The gap between a convincing prototype and a system the business can rely on is made of evaluation, guardrails and ownership — and almost nobody budgets for it.

Next step

Weighing a decision like one of these?

If you are working through a technical choice and want an outside view, get in touch. A straight answer costs you nothing and does not commit you to anything.

We reply to every enquiry within one working day.