When someone in finance needed to know the current exposure across a set of treaty types, it meant submitting a request to the data team, waiting for a response, and hoping the output format matched what was actually needed. If it didn’t - another request, another wait.
Across the organisation, the same pattern repeated daily. Actuaries needed portfolio snapshots to inform pricing decisions. Operations needed status updates across active treaties. The CFO wanted real-time visibility into settlement positions before board meetings. Every question required someone else to extract, format, and deliver the answer.
The data existed. The systems held it. But accessing it required technical skills - SQL queries, database knowledge, familiarity with specific tools - that most business users didn’t have and shouldn’t need.
What we did
We built a natural language querying interface that sits on top of the firm’s structured data layer. Users type questions in plain English - “What’s our total exposure to Treaty X?”, “Show me settlement volumes by quarter for the last two years”, “Which treaties are approaching their compliance limits?” - and get charts, tables, and analysis back. No SQL. No data team requests. No waiting.
Natural language querying only works when the data is clean, structured, and consistently labelled. Because we’d already built the data foundations (standardised interfaces, automated validation, single source of truth), the querying layer could actually deliver reliable answers.
We designed it with guardrails: the system shows its sources, explains its reasoning, and flags confidence levels. Users can drill down into the underlying data. It helps people make better decisions - it doesn’t replace the judgement that experienced professionals bring.
Why it matters
Every reinsurer has data. Very few make it easy for the people who need it to actually access it. The barrier is the data foundation underneath. When your data is clean, structured, and trustworthy, tools like this change how the whole business operates.