Modernising Old‑School Bank Software with On‑Premise AI

Legacy systems in banks are run by retired engineers and rarely change because altering them is costly and risky. A new London‑Budapest startup, founded in 2024 by three former consultants who have seen the same problem repeatedly, is trying to fix this. It raised $6.6 million in a seed round led by VentureFriends, with notable backers including an ex‑xAI co‑founder and a famous football player. The company claims its on‑premise AI platform can read old code written in COBOL, PL/SQL, PowerBuilder and similar languages, pull out hidden business rules, and create living documentation that both humans and AI can use.

What sets the startup apart is that everything runs inside the customer’s own network, not in a public cloud. Banks, insurers and other regulated firms cannot expose their core databases to the internet because of strict compliance rules. By keeping data and processing on‑premise, the platform avoids unpredictable cloud costs and keeps sensitive information in the client’s control. The team also offers a specific tool for modernising Oracle Forms, a technology still common in regulated industries but poorly supported by mainstream tools. They say their migration recipes can speed conversion three times faster and cut documentation work by more than eighty percent, though these numbers are self‑reported.

The long‑term goal is to build “self‑healing” applications where AI agents automatically generate and test fixes, while human engineers only approve them. This vision is still in early stages; the current product focuses on automated discovery and documentation rather than full self‑repair. The startup’s approach highlights the challenge of updating mission‑critical legacy systems while meeting strict security and compliance demands.

Actions