Insights

Notes on data, AI, and insurance — written from inside the work.

Short, specific pieces on entity resolution, insurance AI in practice, enterprise privacy architecture, and lessons from Fabric, Databricks, and Snowflake engagements. No filler, no generic AI takes.

Entity Resolution & MDM Insurance AI in Practice AI Privacy & Governance Platform Engineering
Jul 2026
Entity Resolution & MDM

Why Static Matching Rules Break on Insurance Data

Rules-based MDM was built for data that doesn't change shape. Insurance data changes shape constantly. Here's what that costs, and how machine-learned resolution closes the gap.

Jul 2026
Entity Resolution & MDM

The Economics of Matching: Why Cheap Checks Should Come First

Sending every record through the most expensive resolution method is a common and costly design mistake. Here's how tiered escalation changes the cost curve.

Jul 2026
Platform Engineering

Why Hierarchy Data Needs a Different Engine Than Matching Data

Entity matching and relationship traversal are different computational problems. Forcing one engine to do both slows both down.

Jul 2026
AI Privacy & Governance

Governing the Cost of Agentic AI at Enterprise Scale

Multi-agent AI systems don't stay reliable or affordable by accident. Here's what actually governs cost, behavior, and drift in production.

Jul 2026
AI Privacy & Governance

What "AI-First" Privacy Actually Requires in Insurance

"We don't train on your data" isn't a privacy architecture — it's a sentence in a sales deck. Here's what a real single-tenant boundary looks like, and the questions to ask any vendor claiming one.

Coming soon
Insurance AI in Practice

Underwriting Triage: Where AI Actually Saves Time, and Where It Doesn't

A field-level look at what changes when AI is introduced at first submission review — and the parts of the workflow best left alone.

Have a data or AI question specific to your book of business?

Most of what ends up here starts as a conversation. If you want the direct version instead of the article, ask us.

Talk to us