Focus 03 — Insurance Data Fabric

Can you build the unified semantic layer my analytics, AI, and regulatory reporting all sit on top of?

Why this is hard

Every carrier and large agency has the same problem: book-of-business truth is scattered across Applied Epic / Vertafore / Guidewire / mainframes / Excel, with inconsistent entity resolution between policy, client, location, and exposure. A modern lakehouse (BigQuery / Snowflake / Databricks) with deterministic entity resolution and a semantic layer is the precondition for everything else — analytics, AI, regulatory reporting, and M&A integration.

What I’ve shipped

• SWIA BigQuery production architecture — combined_policy_master_v2 at ~325K rows, book_of_business_v2, attrition_*, client_portfolio_v1, 10× fin_* daily-refreshing views, daily scheduled queries with validation runs. Full case study: Enterprise Data & AI Transformation at SWIA.
• 15-view Power BI suite — finance, attrition, growth, service center, book of business dashboards generated from YAML configs with star-schema relationships.
• 5gVector Coverage Intelligence (private beta) — NFIP flood risk for 694 GA ZIPs + GA DOI bulletins (115 rows) + operator carrier overlays. Isolated subpackage, 133 tests.
• EpicConnect MCP — 18 tools spanning Applied Epic ↔ BigQuery, the canonical example of how an AMS exposes itself as a data-fabric source.
• Pharma Redshift Migration, Pharma Merger Data Integration, Pharma Data Migration — earlier large-scale entity-resolution and semantic-layer work in pharma, transferable to insurance.
• Telecom Vodafone, Telecom Merger — book-of-business consolidation across carriers and country units, same pattern that carriers face during M&A.

What’s coming next

Market signal I track

Guidewire, Duck Creek, Applied Epic, Vertafore, Majesco, EIS, Socotra roadmaps · Snowflake insurance vertical, Databricks Lakehouse for Insurance, BigQuery insurance solutions · IVANS, AMS360, Sagitta, ACORD Forms 25/27/28 · dbt Labs, Atlan, Collibra, Alation, Monte Carlo case studies · NAIC, FEMA, SERFF, state DOI rate filings.

Success metrics

• 5,000+ dataset views/downloads by Q4 2026
• 1 reference customer or insurtech using a public template by Q1 2027
• Speaking slot at one data-platform conference by Q4 2026