Document Base is an internal regulatory document registry for Raiffeisen Bank, a major international banking group. It replaced Lotus, an outdated vendor system unsupported for years, with a modern platform that makes document storage, search, and version control transparent and accessible to all 10,000 bank employees.

Access expanded from ~50 to over 2,500 active users. Manual requests to methodologists dropped ~7 times. Document search time reduced ~1.5x

All regulatory documents lived in Lotus, a vendor system unsupported for over five years. Search only worked through full-text matching. "Filtering" meant toggling columns in a sidebar menu. No clear document statuses, no version history, no way to see who owned what.

Primary users were methodologists, lawyers, auditors, and risk managers. But most bank employees needed documents too, during audits, investigations, or client work. They couldn’t access them independently and had to request everything from methodologists.

For a bank, this wasn’t just inconvenient. It was a compliance risk. The Central Bank requires strict control over document review cycles, and manual tracking meant potential fines.

When I started digging into the existing system and talking to users, I realised the brief undersold the complexity. This wasn’t a UI replacement. It was about untangling years of undocumented business logic.

I framed the problem at two levels:

I interviewed 10 users across roles (methodologists, lawyers, auditors, risk managers) and mapped existing workflows with an analyst.

Research made one thing clear: the system needed to be built around documents, not processes.

Quick low-fidelity concepts, small user groups, simplify based on feedback. I deliberately kept fidelity low because it’s easier to kill rough ideas than defend polished ones. Critical here because every department had its own vocabulary, and only repeated testing revealed which formulations worked.

The registry itself became accessible to all bank employees for viewing and searching documents, something that previously required going through methodologists.

For methodologists, I designed the full document lifecycle within the registry: creating document cards, editing, managing statuses, and exporting data. The document card itself was straightforward, but the creation flow needed careful thought. I broke it into logical steps (core parameters, then responsible persons, then files) so that methodologists who rarely performed this task could manage it without confusion.

I also tested the creation flow with document authors specifically, not just methodologists. I knew that post-MVP, the responsibility for creating document cards would shift primarily to authors, so the flow needed to work for both audiences from the start.

The next phase was approval workflows, a separate but integrated product. I advocated for building from the registry upstream: stable storage first, then creation and approval logic on top.

After that, the priority was separating document types (regulatory docs, forms, committee regulations) into distinct products. Another bank team was already building a reusable semantic search module, and once types were separated and content grew, that module would plug in naturally.

Investing in business logic before designing. The framework chain (JTBD to OOUX to IA to Flows) kept every decision traceable and gave me a shared language with engineering and business.

Push harder on document type separation. I spotted the structural differences during research but accepted the decision to defer. Post-launch, this created confusion because users expected different types to behave differently.

The real complexity wasn’t in the interface. It was in translating poorly documented business logic into something that felt simple, and aligning the team around what to build first