AI can help organisations analyse legacy systems faster
One of the biggest barriers to modernising legacy technology is understanding how existing systems work.
To modernise systems, organisations first need a clear picture of how applications work, connect and depend on one another. That process can take weeks or even months. Documentation is often incomplete or out of date, knowledge may be concentrated in a small number of specialists, and some systems rely on technologies that few engineers still work with regularly.
Our pilots indicate that AI-assisted approaches could help technical teams analyse some legacy systems more quickly.
Across participating organisations, activities estimated to take weeks using traditional approaches were completed in hours.
Working with organisations including the Student Loans Company (SLC), the UK Hydrographic Office (UKHO), the Met Office, and our delivery partner Kainos, we found that AI can help teams analyse unfamiliar systems, identify dependencies and build technical knowledge more quickly than traditional approaches alone. The pilots covered different systems and technical environments, so the results should not be treated as directly applicable to every legacy estate.
However, for organisations managing large and complex technology estates, this could make some modernisation activities easier to prioritise, plan and deliver.
What happened at the Student Loans Company
At the Student Loans Company, teams explored how AI assisted approaches could be used to analyse an estate of 33 applications written in Clojure and support future modernisation planning.
Using traditional approaches, the work was estimated to take around five weeks. During the pilot, teams completed the analysis in less than a day.
The analysis gave the team a clearer view of the 33 applications and their dependencies, which it used to inform migration planning and work towards supported versions of Java.
The most significant finding was not the reduction in effort itself, but the ability to unlock work that had previously been difficult to prioritise.
What happened at the UK Hydrographic Office
At the UK Hydrographic Office, teams focused on a 20-year old desktop application that previously had no automated tests.
AI assisted techniques were used to generate automated tests for the application. Test coverage increased from 0% to 96%, while regression testing that had previously required around 15 days of manual effort was reduced to approximately 30 minutes.
The benefits extended beyond faster testing.
Teams reported greater confidence in making changes to the application safely and maintaining it over time. For many legacy systems, the perceived risk of change can be a barrier to improvement. The pilot indicated that AI-assisted testing could give engineers more evidence when assessing proposed changes.
Why this matters
The examples from SLC and UKHO illustrate a wider challenge facing organisations across the public sector.
Legacy systems often support critical services, but analysing them can be complex and expensive. Teams may spend weeks mapping dependencies, reviewing code and building technical knowledge before they can assess risks or plan change.
The strongest finding from the pilots was that AI can help accelerate this discovery and assessment phase. That may sound modest, but it addresses a critical point where many modernisation efforts slow down or stall.
By helping teams understand systems more quickly, AI assisted approaches can make it easier to:
- assess risks
- identify system dependencies
- prioritise investment
- improve technical documentation
- develop modernisation plans
In short, AI can help organisations spend less time understanding their systems and more time improving them.
AI is not a substitute for good delivery
The pilots also reinforced an important lesson.
AI does not remove the need for good governance, engineering expertise or delivery capability.
Successful outcomes depended on organisations having the right foundations in place, including clear ownership, access to systems and information, experienced technical teams and established delivery processes.
In practice, the pilots showed that AI amplified existing strengths and weaknesses rather than replacing them.
Used well, AI can accelerate analysis, improve visibility of complex systems and help organisations make progress more quickly. Used without appropriate controls, it can amplify existing problems and accelerate poor decision making.
Where AI-assisted analysis may help
The pilots suggest that AI can significantly reduce the effort required to assess complex systems, helping organisations move more quickly from discovery to modernisation.
For organisations managing large and complex technology estates, this can help engineers, architects and delivery teams to identify risks earlier, making modernisation activities easier to plan, justify and deliver.
What happens next
We are continuing to work with organisations using AI assisted approaches to analyse, assess and modernise legacy systems, building on the lessons from these pilots.
Alongside this work, we are developing a playbook of good practice that captures practical approaches, lessons learned and implementation considerations for organisations looking to apply AI to legacy technology challenges. The playbook is currently undergoing user testing, review and approval ahead of publication.
Get involved
If your organisation is exploring AI assisted approaches to analysing, documenting or modernising legacy systems, or would like to share experiences and engage with our wider Firmer Foundations programme, we would like to hear from you. Get in touch at firmer.foundations@dsit.gov.uk
Leave a comment