Years of legacy logic, 
decoded by AI in days

Overview

Modernising a system nobody fully understands usually means months of discovery before a line of new code is written.

For a FTSE 100 group whose forecasting platform decides where to allocate multi-million-pound budgets, an integrated AI toolchain compressed that discovery from weeks to days. 

The successor platform went live with higher forecast accuracy and stronger adoption than the one it replaced, its design no longer locked away in a handful of engineers' minds.

25%

of the project timeline and spend reclaimed

Weeks → Days

from weeks of analysis to writing code in days

100%

every user chose to switch

Before

A system nobody could fully explain

Documentation described what the system did, but why it had been built that way had never been written down. That reasoning survived only as a handful of engineers' working knowledge. Modernising meant preserving that hard-won logic without reintroducing the slow response times and uneven accuracy users already disliked. The conventional fix, a dedicated architect and 8 to 12 weeks of discovery, would burn a quarter of the timeline before a line of new code was written.

After

Years of legacy logic, decoded in days

The new platform went live with roughly a quarter of the project timeline reclaimed, shipping with targeted algorithms matched to each data type, higher forecast accuracy than its predecessor, and migration so low-friction that legacy users moved across voluntarily rather than being forced.That speed came from the method. What usually takes weeks with a dedicated architect, the team did in days: AI ingested the documentation, mapped the existing codebase, and read the transcripts explaining why past decisions were made. Each accuracy change was validated with the client's data scientists, so gains were evidenced, not asserted. The system that was a liability to maintain is now an asset to build on.