Confidential InsurTech
A confidential insurance technology partner needed to turn a tangled web of entity relationships and unstructured policy documents into a single, queryable system of record. We built the platform in six months.
The Bet
Commercial insurance runs on relationships: the carriers, brokers, program managers, and insured entities that connect to one another in layered, constantly shifting ways. For our partner, an insurance technology company, that web lived in spreadsheets, documents, and people’s heads. The bet was to turn it into a single system of record that could map every entity relationship and read the coverage behind it.
The Complexity
Two hard problems at once. First, the entity graph: modeling how organizations relate up and down the insurance chain, where the same company can be a counterparty in one context and an insured in another. Second, the documents: coverage information arrived as unstructured files across lines like errors and omissions, cyber, and crime, with no consistent schema to extract from.
The Thinking
We modeled the entity graph first, then layered coverage intelligence on top. Rather than force a rigid schema onto a messy domain, we designed a structure that could represent upstream and downstream relationships natively, then used AI to extract structured metadata from documents and attach it to the right entities.
The Build
Over six months we shipped the platform in phases: the core entity-relationship model, enablement of downstream entities, upstream relationship mapping, and a metadata-extraction layer that read coverage details across multiple lines. The work moved at a steady cadence, 23 deliveries with an average cycle time under seven days.
The Approach
This is the kind of problem Nolte is built for: an ambiguous, document-heavy domain with no clean data model to start from, where the hard part is judgment about what to build before writing code. We scoped it into phases, derisked the entity model up front, and delivered against a plan the business could follow. That is what it means to innovate predictably.
