Want to know more?
Are you interested in this project? Or do you have one just like it? Get in touch. We’d love to tell you more about it.
How a financial services giant gained insight into legacy systems with generative AI
How a 3-week experiment helped our client understand how AI could unlock legacy code and speed up future development
Our client, a leading global insurance brand, needed support to launch a business-critical initiative to white label its digital motor insurance for use by other financial services companies. We provided delivery and enablement services to modernise the customer’s technology stack, including cloud-based build-test-deploy pipelines and automated testing. Despite navigating a 15-million-line .NET monolith and rigid manual security processes, the client successfully launched the service in May 2025. This partnership has since scaled significantly, driving high-volume daily sales and establishing a blueprint for their future digital transformation.
Outcomes
2.5 days
to extract more information about a legacy system than previous 4-week efforts
3.5 days
to complete the first phase of a shelved frontend rebuild, versus the original 2–4 week estimate that made the work economically unfeasible
3 weeks
sprint completed early, upskilling principal engineers in advanced AI techniques through pairing — meeting capability goals ahead of schedule
About the client
Our customer is a global leader in insurance and asset management, dedicated to protecting people and businesses. The company’s UK retail business focuses on providing innovative digital insurance products to millions of customers across the UK.
Industry
Financial Services
Organisation size
150,000+ employees
Location
Global
Project length
3 weeks
Challenges
Decades of legacy debt, limited documentation, and the cost of standing still
Like many established global finance brands, our customer has multiple legacy systems that date back decades, with limited documentation and access to the skills to manage and update those ageing systems.
The company’s tech infrastructure includes a mixture of .Net, Java and Python applications and mainframe systems dating back decades. The scale and criticality of these platforms mean that modernisation initiatives require significant time, budget and specialist availability, often competing with other delivery priorities and slowing the pace at which change can be introduced, with projects often put on hold because of the time and cost involved in migrating services off these legacy systems.
The customer’s UK team have been considering how GenAI could help their team for some time, particularly around understanding and modernising legacy systems. They also hoped that GenAI could help their teams prototype new solutions that might help the company justify re-starting some of the work that had previously been put on hold.
Solution
A 3-week paired assessment to answer three critical questions
The customer approached us to help gain insight into how AI could help to address these challenges.
We designed a 3-week assessment to evaluate up to three legacy systems using a paired model where our consultants worked directly with the company’s own principal engineers. The assessment aimed to answer three questions:
- Can AI tools help teams understand poorly understood legacy systems faster?
- Would a small team achieve more with AI tooling than without?
- What can we prototype in one week to reassess deprioritised work?
The assessment started by evaluating a highly complex .Net application for insurance product configuration. We used GitHub Copilot integrated with Claude, using meta-prompting and proven engineering analysis techniques to understand the legacy system’s logic, and to generate new documentation.
Next, we helped our customer to reevaluate the feasibility of using AI to rewrite an existing user journey – specifically change of address – using React. Previously, this had been put on hold because of the time and cost involved, but using Copilot and Claude as a pair programmer allowed us to make a copy of the existing journey, break it down and build a new proof-of-concept.
Results
Faster understanding, faster delivery, and a team ready to continue independently
The assessment demonstrated that generative AI can help teams understand legacy systems faster, prototype sooner and re-evaluate previously deferred work.
The results were powerful and immediate. The customer’s Head of Engineering remarked that the Equal Experts team ‘extracted more information in 2.5 days than we did in four weeks.”
The proof-of-concept we built for the address-change journey had originally been expected to run for 2-4 weeks. Using GenAI we were able to deliver 50% of the final product in just 3.5 days.
The assessment also helped the customer to build confidence among its in-house engineering team. The organisation’s engineers met capacity-building goals early enough that the Head of Engineering is confident that their engineers could continue using these tools independently.
Key lessons learnt about GenAI during the engagement
- AI speeds up understanding. It helped the team decode a legacy system quickly, and unlocked work previously considered too difficult or time-intensive.
- A small team using AI might be more productive. Small teams might be able to complete work in days rather than weeks.
- The approach showed value to roles beyond engineering. Analysts, testers and product teams could all benefit from an AI-enabled approach.
- Faster learning flattens the learning curve. Capturing, sharing and applying knowledge in real-time meant teams spend less time getting up to speed and more time delivering value.
While bottlenecks can remain a challenge in any large engineering team, this programme helped to clearly demonstrate the power of AI to amplify skills and increase productivity.
Equal Experts is continuing to help the customer to scale these practices across the wider engineering team.