GenAIJul 2026
Why enterprise GenAI stalls after the pilot
The blocker is rarely the model. It is evaluation, access control and the absence of an audit story.
Insights
Field notes on production AI
Practical writing on retrieval architecture, agentic workflows, healthcare AI and claims automation — drawn from live enterprise programmes.
RAG best practicesJun 2026
Chunking is a policy decision, not a parameter
How document structure in policy wordings should drive your retrieval strategy — and what breaks when it does not.
Agentic AIJun 2026
Five agentic use cases that actually survive risk review
Bounded autonomy, human checkpoints and reversible actions: the pattern behind agent deployments that ship.
Healthcare AIMay 2026
Pre-authorization is the highest-ROI AI use case in health insurance
Volume, structure and clear ground truth make pre-auth the best first target for an explainable copilot.
Claims automationApr 2026
Measuring claims AI: the four numbers that matter
Handling time, straight-through rate, decision consistency and audit exceptions — everything else is vanity.
Book chapterMar 2026
The Autonomous Financial Layer — Chapter 2 excerpt
Retrieval as institutional memory: why the vector layer, not the model, is the durable asset.
Want these in your inbox?
Reach out and I will add you to the occasional note on production AI in insurance and banking.