Beyond financial services
Most of our work is in regulated finance, but the conditions we are brought in for are not unique to it. Wherever a decision is expensive to reverse, and wherever the value of the business rests on something a machine should not be left to settle on its own, the work is recognisably the same.
SECTORLuxury goods
The longest feedback loop of any sector we work in
We work with whisky brands, where filling a cask this year means finding out whether you were right in eighteen. It is the most literal version of a decision you cannot unmake, and it makes every assumption in a forecast a commitment rather than an estimate.
The value of these brands also rests on provenance, craft and restraint, which are the first qualities automation erodes if nobody has decided where it is not allowed to go. That decision is the work.
Where automation is safe, and where it is not. Separating the operational load nobody is paying for from the craft and provenance they are.
Long-cycle stock and inventory judgement. Decisions whose consequences land decades later, made on evidence rather than on last year's pattern.
Direct-to-consumer and collector experience. Reaching buyers directly without cheapening what they are buying into.
Marketing and content operations inside strict codes. Alcohol advertising rules, age statements and geographical designations apply to AI-generated material exactly as they apply to everything else.
SECTORLife sciences
Where the evidence has to survive an inspection years later
A decision has to be defensible to an inspector long after the people who made it have moved on. Validated systems, traceable evidence and documented human oversight are not preferences here, they are conditions of operating.
Evidence and documentation load. Scientific and clinical experts spending their time assembling and formatting evidence rather than interpreting it.
Validated by design. Capturing requirements so that a system's behaviour can be traced back to intent, which is what a validation exercise is asking for in the first place.
Where human oversight must stay. Scored by consequence and reversibility rather than left to habit, and written into the architecture rather than into a policy document.
SECTOREarly-stage ventures and research commercialisation
Designing the business before the legacy exists
Knowing what your AI-powered startup solves, and for who, is the key difference between a good technology and a great business. We work with the University of Edinburgh's AI Accelerator, helping founders do the thinking most startups skip.
We help them ask the right questions before there’s any process debt to work around.
Futures thinking. Understanding where your customers and industry are heading so you can design an informed business strategy, before early product decisions harden into constraints.
Business model design. Where AI changes the unit economics, and where founders are assuming it does when it does not.
Implementing AI in a small team. No platform, no data estate and no governance function yet, which changes what is sensible rather than removing the need to decide.
Defensibility. What this venture will know that a competitor with the same models cannot. The question most early AI businesses answer far too late.
“
The problems we solve for regulated banks aren't unique to regulated banks. Any organisation making decisions that matter – to a customer, an employee, a market – needs the same discipline: pause, consider, then act. That's not a compliance requirement. It's just good design.
Jamie Spratt, Principal Consultant
