We helped our banking client manage projects more quickly, accurately, and with less risk using our AI-based LLM Wiki solution.
Sector
Service & Technology
The Challenge
Our client is one of Hungary’s top banks, where we participated in a data migration project.
Our experts involved in the project had to quickly gain an understanding of the structure and operation of the system in question, even though the necessary information was scattered across various business and technical specifications, database descriptions, and developer documentation:
- business operations in the business specifications,
- technical implementation in the technical specifications,
- data structures in data models and database schemas,
- and the actual implementation in Java/SQL code and GitHub repositories.
This is a challenge because simply reading a single document is not enough to understand a new topic. You need to be able to see what data a business process involves, which technical components it is linked to, which database tables it appears in, what impact it has on reporting, and what risks may arise in connection with it.
The Solution
In response to this challenge, we created our AI-based LLM Wiki solution, which transforms existing project documentation into a unified, searchable, and interconnected knowledge base. The LLM Wiki processes the submitted source materials and organizes them into a wiki-like structure, separating, for example,
- business processes,
- data domains,
- data models,
- entities,
- technical components,
- issues,
- risks,
- summary of source documents,
- and related methodological descriptions.
The result is a structured knowledge base: for any given topic, it becomes clear which documents the information comes from, what other concepts or components it is linked to, what risks have arisen, and where further details can be found.
Technology
GitHub Copilot / VS Code
The Result
Thanks to the completed LLM Wiki,
- time can be saved, since each topic can be reviewed much more quickly;
- risks can be identified earlier, based on the specifications alone, saving time that would otherwise be spent on incorrect development and testing;
- business rules and technical implementation can be better aligned;
- and finally, data model and architecture analysis also become easier.
