THE CHALLENGE
Many financial institutions are encumbered with legacy systems. While these systems are reliable, they present a plethora of risks to business. From the lack of a work force skilled in operating the systems and operational bottlenecks, to the urgent need to modernise and remain compliant with new regulations, financial institutions have a strong incentive to find workarounds.
Chief among these is a solution that translates and understands legacy codebases, enabling modernisation whilst avoiding the need to rely on scarce legacy tech expertise. Existing COBOL-to-Java often generate unreadable ‘Jobol’ language, therefore necessitating tools that offer code explanation as opposed to direct conversion.
THE SOLUTION
Critical Software has developed a solution powered by an AI generative chatbot, designed specifically to analyse and explain legacy codebases. This will enable financial institutions to:
- Analyse legacy programmes and provide detailed explanations of functionality and business logic
- Generate pseudo-code to give dev teams a strong foundation for rewriting code in modern languages (e.g. Python or Java)
- Understand data models and architectures, offering insights into how legacy applications interact with databases
- Create code-flow diagrams to support modernisation processes
- Receive real-time feedback on questions regarding the legacy codebase
THE RESULTS
The AI-powered chatbot has so far delivered positive outcomes, including:
- Rapid conversion was achieved for Java developers, with over 400 lines of COBOL routines translated into Java within a one week period
- Testing was completed with a success rate of 80%, with 300 test cases passing on the first attempt. The high accuracy level is pivotal for reducing the need for extensive debugging or reworking
- Improved client satisfaction
Technical benefits of the chatbot include:
- Faster time-to-market for updated platforms
- A deeper understanding of business rules, preserving and optimising critical business logic
- Notable cost savings
THE TECHNOLOGY
- Large language models (LLMs)
- Retrieval-augmented generation (RAG)
- Interactive AI interfaces
- Cloud-first architecture
- Private LLM support