Legacy system modernisation

Legacy systems hold years of policy, workarounds and decisions that were never written down. That logic sits in places that conventional discovery struggles to reach:

• Source code, stored procedures and batch processes
• Spreadsheets and macro-enabled workbooks
• Policy documents, manuals and the knowledge of long-serving staff
• Systems where the original supplier, or the source code, is no longer available

So, modernisation keeps getting deferred. The old system costs more to run each year, the people who understand it retire, and the risk grows quietly. Across Scotland’s public sector, around £3.80 goes on running existing technology for every £1 spent on changing it.

AI changes the economics, but only when it is applied with discipline. We use AI to do the slow, expensive work of discovery, specification and build with far less effort. Every finding then goes to people who know the service before anything is built. You start with evidence, not with a target platform.

Most AI tools help an individual work faster. We have gone further and turned Storm ID’s whole delivery method into a system that runs the same way on every project.

• Agents do the work. We have built eight AI agents, each trained for one role on a delivery team: business analyst, user researcher, UX designer, content designer, tech lead, two architects and a test writer. Each agent has written instructions for its role, like a job description and a playbook combined.
• Skills hold our know-how. We have written down 40 pieces of specialist knowledge that the agents draw on. These include how we write .NET and Node code, how to build prototypes to Scottish Government and GOV.UK design standards, how to write secure code, and how to test a service so nothing breaks when it changes. Each project switches on only the skills its technology needs.
• Gates keep quality high. Every project moves through the same stages: Discovery, Prototype, and Build and Test. Each stage has a gate. Before work can pass, the required outputs must exist, automated checks must pass and a named person must approve them.
• How it works in practice. Say the Discovery Accelerator finds an eligibility rule hidden in an old spreadsheet macro. The business analyst agent writes it up as a plain-English requirement and links it back to the cell it came from. Your service owner confirms the rule is still correct. Later, the test agent writes a test that proves the new service applies the rule in exactly the same way. A person signs off at every step, and you can trace the rule from old system to new.
• Why this is valuable. Because the method is written down, the quality of your project does not depend on which people we put on it. Every project follows the same standards and checks. And because it is shared across all our work, every improvement we make on one engagement is ready for the next.

• AI output is treated as a draft until people confirm it, and a person approves every output at every gate
• Independent reviewers check prototype fidelity and code quality before work moves on
• UK government standards checks and a secure coding guard run as part of the method, not as an afterthought
• Sensitive material stays under your control, and data is minimised before any analysis
• Analysis runs in approved environments with identity controls, MFA and least-privilege access
• Model providers retain none of your data
• AI is used to build the service, not as something the live service depends on
• Delivered under ISO 27001 and ISO 9001 certified processes

• Standing still keeps getting more expensive. Most legacy budgets go on keeping systems running, not on improving them.
• Specialist knowledge is leaving. The people who understand older systems are retiring faster than they can be replaced.
• AI has changed the payback. In our modelling, payback on a legacy system costing £1m a year falls from about 10.5 years to 6.3 years.
• Reform depends on it. Better services and lasting savings need systems that can change.

Storm ID helps you move from deferral to delivery. You start with one system and measure the results before you scale.