A boutique AI studio. Small on purpose.
An AI implementation and enablement studio. We are AI-user specialists — not the people who build the models, the people who work out what they are good for in a real business, and then get everyone there using them.
The short version
This technology is the most interesting thing to arrive in our working lives, and we enjoy it enormously. Most of what we know came from using it every day on real work until we found where it holds up and where it falls over. Sharing that is the job. We would rather show someone in an hour what took us a year to learn than keep it to ourselves. Part of that thinking is written up in Prompt engineering is the wrong name.
Over 15 years in the problem, on the business side of it: master data governance, SAP, financial analysis, data migration, workflow automation and business intelligence. AI in the form people now mean by it did not exist before ChatGPT arrived in 2022, so nobody was using it, us included. We started building with it close to day one and have not stopped.
Two things surprise people every time. The first is how much these models can already do. Most people have seen one write an email and concluded that is the ceiling, when the same tool will read three hundred contracts, reconcile a year of statements, or build the thing you were about to hire a developer for. The second is how close it all is. There is no enormous budget or specialist team standing between you and this. The tools are cheap, they open in a browser tab, and the real gap is that nobody has shown you what to point them at.
Where that came from
From the largest businesses to the smallest: Kontoor Brands, Helly Hansen and Wrangler jeans, Morgan Stanley, Accenture, Cambridge University, Vistra, Getronics, and earlier, Starbucks, Connells, Brio Software and SeeBeyond. Education, real estate, construction, logistics, consulting, apparel, retail, investment banking, corporate services and software, in China, the UK, the EU, the USA and South Africa. We have run and worked in remote teams across Asia, the EU and Africa.
That range is the useful part. A problem that looks novel in one industry is usually one we have watched somebody solve in another, and we can tell fairly quickly which it is.
Two things. The second one has two halves.
We build with AI, and we build AI in
Two different things, and both matter. We use AI to do the building, which is why the work comes back faster and costs less than it used to. And where the job calls for it we put working AI inside what we build, so the finished thing keeps doing the clever part long after we have gone. Tools, workflows, dashboards, assistants, document pipelines — the catalogue of 269 jobs is the honest list of what we take on.
We get your people using it
Tailored to the organisation and to the individuals in it. Where it helps we start with a survey or interviews, so the sessions answer what is in the way rather than what a generic course assumes.
We work with the lid off
We are generous with what we know: which tools, why those ones, where they fall over, and how to run the thing yourself afterwards. That costs us nothing, because the tools change every month and the list was never the valuable part. What is left when you give the knowledge away is judgement and attention — knowing which model to reach for, where each one quietly fails, when a cheap one will do and when it absolutely will not. That comes from years of using them and it goes stale too fast to write down. The knowledge is free. The attention is what you are paying for.
Software as an asset, not a subscription
Whatever we build, you own outright. It is yours, not licensed back to you. And we teach you how to change it, so it grows with the business instead of freezing on the day we leave. The alternative is another monthly subscription for something you cannot alter, that may not be supported properly, and that costs more every year.
Security shaped around your data, not ours
Every client gets a setup that matches how sensitive their material is. Where it matters, nothing leaves your network: we run open-source models on your own infrastructure so the data never reaches a third party. Where that is more than the job needs, we work on masked or redacted copies instead. We agree which of those applies before any real data moves.
Human judgement on the AI output
AI does the heavy lifting, then a person applies human judgement to what it produced. Every output gets read by someone who knows the business it is going into: is this right, is it complete, and has the model flattened the nuance that mattered? Models are confident when they are wrong and they miss the things that are obvious to anyone who has done the job. Catching that is judgement, not a checklist, and someone has to own it.
A small team, and a handover protocol
We are deliberately small: a core of people, a remote team of interns, and a set of AI agents and bots we have trained and keep training, running on whichever models suit the job. Because we are small, there is a written handover arrangement with a named successor, so work in progress does not disappear if something happens to any one of us. We will walk you through it on request.
We are picky, and it is not about size.
Being boutique means taking on a small number of clients at a time and giving each one real attention. So we choose carefully, and what we are choosing for is fit, not headcount. A founder working alone can be a better job than a large company, and often is.
We work with chief executives of corporates, with founders and entrepreneurs, and with their staff and teams.
Organisations of any size — startups deciding what to build, established businesses carrying fifteen years of accumulated mess, private family offices and the businesses they hold. What matters is that someone senior wants this to work.
Individuals, privately. Executives and CEOs engage us directly, at an individual fee, outside their organisation and in confidence — for coaching on the technology itself, or to think through a business concept with someone who has no stake in the internal politics. Those conversations stay between us.
Anywhere the working language is English. We work remotely across time zones and have done for years.
Want to work on this with us?
Not a full-time job — a few hours a week or a few a day, fully remote, on a partnership basis. It works well alongside a career you already have.
See careers →