KR&A
We use AI to read, map and rebuild legacy estates, running open-weight models on infrastructure you control. Your code never leaves your jurisdiction, and every change goes through your existing release process.
less manual patching effort for a global payments network, with every change approved through the client’s own CI/CD process.
The codebase read end to end, and what it does explained in business terms.
Dependencies, risks and dead code mapped before a line is changed.
Module by module, highest cost and risk first rather than oldest first.
New behaviour proven equivalent to old before anything reaches production.
We do not. Modernisation is where our three disciplines meet: a method that rebuilds, an architecture that keeps the work in your jurisdiction, and a record that proves it was done properly.
KR&A
AI is the primary implementer, not an autocomplete. When the machine writes the code, the code stops being the record of what the business decided. We keep people on the decisions and the knowledge somewhere people can still read it.
Sovereignty is architectural, not contractual. A European region on a hyperscaler invoice is not control. We build on open source and open weights wherever possible, and hold the remainder to jurisdictional guarantees you can enforce.
Every system encodes a distribution of value, whether anyone chose it or not. We make the choice explicit. Who it is built for, who is paid for it, and who is able to use it.
Three convictions, one practice. We do not run them as separate service lines, because our clients do not meet them as separate problems. A system built faster than it can be specified, running on infrastructure you cannot leave, trained on data nobody can account for, is one failure with three names.
We work in four modes across all three. We assess, so you know where you stand before anyone commits to anything. We train, because a method that does not transfer is just another dependency. We build, alongside your engineers rather than instead of them. We govern, so the position still holds after we have gone.
Systems the business depends on and nobody can safely change. We read them, map what the business relies on, and rebuild them module by module, proving each one behaves as the old one did before it reaches production.
For business critical systems, the question is not how much a machine can write. It is where a person has to decide, review and sign. We put those control points in the architecture rather than leaving them to individual discipline.
Generated code is disposable. What the business agreed is not. We hold that agreement in behaviour specifications, in interface contracts written before implementation, and in tests that read as statements of intent, so the organisation still knows what it built and why.
Practitioner training and co-delivery. Your team is leading the work by the end of it.
Every component of your toolchain measured against one test. Forkable under an open licence, and jurisdictional control of data flow at runtime. Vendor domicile does not pass it.
A working European stack. Git forge, project management, communications, models and orchestration. We run our own business on it.
Off proprietary platforms, with the exit cost established before you commit rather than discovered when you leave.
An unannounced vendor model update is a re-certification event. With a proprietary model, your compliance posture is set by someone else. Open weights return that decision to you.
The European Accessibility Act has applied since June 2025 and is now being enforced. Most digital estates fail it. We find where, fix what matters, and leave the standard in your definition of done rather than in a remediation backlog.
Where the material came from, who is owed for it, and how the system proves both. The Distributed Equity Licence applied in practice.
An AI-native delivery method changes the shape of an engineering organisation. Which roles move, which disappear, and what a fair transition looks like. Nobody selling the method wants to answer this. We put it in the scope.
Chapter I, universal scope, in force since February 2025. It applies whatever your risk classification, which makes it the first obligation rather than the last. We assess the position and deliver the training that discharges it, at practitioner level rather than awareness level.
An assessment is the cheapest way for both of us to find out whether we are the right people for the work.
A fixed-scope baseline of your legacy estate, produced on your infrastructure. You keep the findings whether or not we do the work.
Fixed scopeRuns on your infrastructureThe findings are yours
What each system does, written for the people who own it rather than only the people who maintain it.
Dependencies, undocumented logic and single points of knowledge, ranked by exposure.
Which modules to modernise first, ordered by maintenance cost and risk rather than by age.
Which tools and models the programme would rely on, tested against licence and runtime control.
Thirty years inside enterprise estates, from CERN to NatWest. The first conversation costs you nothing.
Book a Discovery Call