A written record of what your system does
Its rules and dependencies, each linked back to the source code.
It was written decades ago, in COBOL, PL/I, Fortran or another language few people still write. It is expensive to run and hard to change. The people who built it have retired or left, and the documents no longer match what it actually does.
You rely on it every day, so you cannot turn it off. It gets harder and more expensive to keep running every year. And if a new system gets even one rule wrong, your customers feel it.
Replacing an old system should not put your business at risk.
£48.65 millionthe fine after one platform switch
In April 2018 TSB moved its customers onto a new banking platform. It failed on the day, and a large share of its 5.2 million customers were affected. The regulators later fined TSB £48.65 million, and the former chief information officer was fined personally.
FCA and Bank of England, 2022 and 2023. FCA notice, Bank of England notice, PRA decision.
Sovereign finds every difference before you switch.
Who it is for
Stock, order and warehouse systems on older midrange computers.
Production, stock and delivery systems that run every shift and were built decades ago.
Tax, benefits and licensing systems that depend on older code and fewer people who know it.
Patient administration and hospital systems that cannot go offline.
Billing, metering and asset systems that serve whole regions.
Billing, scheduling and network systems built decades ago.
Core banking, policy and pension systems running on decades-old mainframe code.
Systems that must be modernised inside sealed, secure networks.
After Sovereign
A worked example

We gave Sovereign the source code of a public mainframe credit card application. It mapped all 44 programs, 62 shared data definitions and 55 batch jobs, and found a missing file before any rebuilding started. Sovereign rebuilds and proves COBOL, PL/I, JCL, Fortran, MUMPS, REXX, Perl and ColdFusion, each checked against its own compiler or interpreter.
Old and new run on the same inputs. Each case lights up as it matches. When one differs, the report shows the cause, the fix and the re-run, with the evidence kept for your team.
The Sovereign method
We work out exactly what your system does: its rules, data and dependencies. Each rule points back to the lines of code it came from, so your team can check it.
We build the modern replacement from that record.
We run old and new on the same inputs and show you every match and every difference, until the report is clean.
When a part is proven, the new system takes over its work in small steps while the old one stays ready. If results drift, the work goes straight back to the old system. You approve every step.
Where it runs
Sovereign comes as a sealed package that installs on a machine inside your own network.
It is built to run with no internet connection, including in air-gapped sites. Reading, rebuilding and proving do not need any outside service.
Reading, rebuilding and proving use no AI model, so every result is exact and repeatable. Nothing leaves your network without your say.
Our promises
What you receive
Its rules and dependencies, each linked back to the source code.
Modern code for the agreed part, built from the record.
What matched, what differed, what is still open, and instructions to re-run it.
How to start
Before anything changes, we map your system: what you have, what it depends on, and where to start. Scope agreed in writing before we begin.
The old system able to run, representative test cases, and one person who owns the decision.
For each person in the decision
Two to four weeks for one part of your system, once access is ready.
Every difference is shown to you with the inputs that caused it and the lines of old code behind it. Some differences are faults in the old system that you choose to keep or fix. Some are gaps in the new one. You decide which, and the proof is re-run until the report is clean.
We agree with you how each part will be proven before the pilot starts.
Where you say. We agree secure access in writing before you share anything: your environment or a sealed one we set up, named people only, and access that ends when the work ends.
You do. The written record, the new code and the proof report are yours, and your team can re-run the proof without us.
Each step is fixed in scope and agreed in writing before work starts. Price is set out in the written scope, so there are no surprises later.
Email us about the system and what worries you about changing it. We reply with a proposed first step.