You've been pitched AI before. It probably didn't work.

You've had the pitch before. Someone with a deck full of logos tells you AI is about to transform your business, shows you a chatbot demo, and asks for a number with too many zeros on it. Maybe you said yes once. Maybe you're still waiting for it to do anything useful. Maybe the invoice arrived well before the value did, if it ever did.

A lot of those projects never started with a business problem. They started with pressure. A board member who'd read an article. A competitor who'd announced something. A feeling that everyone else was already doing it, so you'd better be seen doing it too. Roll out Microsoft Copilot to the whole company and hope people find a use for it. That works fine if your organisation is full of self-starters with a strong technical instinct and middle managers actively steering the adoption somewhere useful. Most organisations aren't built like that. Most of those licences just sit there, technically deployed and practically ignored, while someone in finance quietly wonders what they're paying for.

We don't start with the technology. We start with the number.

We only take on work where you and we have identified, together, something specific worth doing, with a return you could reasonably expect to earn back within six months of delivery. Not a projection built on industry benchmarks. Your numbers, from your business.

If that number doesn't exist yet, that's a legitimate answer too. It means the honest next step is finding it, not starting a project that adds to your list of expensive AI things that didn't stick.

AI is one tool in the box, not the point of the engagement.

Most of the time, AI ends up being the smallest part of what actually fixes a business process. We look at how the whole thing runs end to end, not just where a chatbot could be bolted on. Sometimes the fastest win is a piece of software you already own and aren't using properly. Sometimes it's a form that shouldn't exist. And almost always, part of the fix is the people in the process. New tooling dropped into an old way of working just gives you a faster version of the old way of working. The engagement isn't finished until the human side of the process has been rebuilt around the new way of doing things, not layered on top of the old one.

What this actually looks like, inside a real business

Take a construction contractor we're currently working with in Europe. The brief that brought us in was admin overload: an office team buried, and an owner assuming the fix was faster software for the back office.

We didn't start there. We started by asking whether the admin team was actually doing administration, or spending their week compensating for information that arrived from site late, incomplete, on paper, or not at all. Timesheets scrawled and half-legible. Site diaries written from memory three days later, if they got written at all. A verbal agreement about extra work on-site that never made it into anything chargeable. Individually, each of these looks like a small annoyance. Someone just has to chase it up. Added together, it's most of what makes the job exhausting.

So we ran the arithmetic instead of guessing. Ten people on site losing 15 minutes a week untangling their hours, plus a similar amount of time on the office side collecting it, chasing it, and retyping it. Across a working year, that's around 230 hours, and close to €5,000, on timesheets alone, on a ten-person firm. We found six more workflows carrying a similar hidden cost. Nobody had ever added it up before, because on its own, every individual piece looked too small to bother with.

That's the number before a single line of automation gets built, and it's the number we hold the client to. We're inside this business now, working workflow by workflow, and we're building the case study in public as it happens rather than polishing one after the fact. If it works the way the arithmetic says it should, you'll see the real figure here, not a rounded-up one.

We've done this kind of work since long before it involved AI.

We're not a new AI shop staffed by people who fell in love with a model and went looking for a market. We've spent our careers leading business and process change, across industries, long before "AI strategy" was a job title anyone printed on a business card. The tools change. The discipline of finding where the value actually is doesn't.

If we can't point to the number a piece of AI is supposed to move, we won't sell it to you. Sometimes that number points somewhere else entirely, and we'll tell you that instead of building it anyway.