Insights & field notes
Thinking clearly about applied AI.
Practical notes on document intelligence, regulated onboarding, and the engineering choices that make automation hold up in production.
Talk to an architectNo hype cycles · Just useful patterns
Latest thinking
Read the room, then read the system.
7 min read
Confidence is a product feature in document AI
May 14, 2025
Why confidence scores, evidence links, and clear exception queues matter as much as raw extraction accuracy.
6 min read
Designing onboarding flows for the exception, not the average
Apr 29, 2025
A practical framework for building regulated onboarding journeys that make edge cases feel intentional.
9 min read
RAG systems with an audit trail
Apr 08, 2025
The architecture patterns we use when every generated answer needs a source, a timestamp, and an owner.
5 min read
Human-in-the-loop is an operating model
Mar 21, 2025
Moving beyond a checkbox: how reviewers, models, and policy teams share responsibility in production.
8 min read
How to measure document automation beyond straight-through rate
Feb 27, 2025
A scorecard for accuracy, review effort, latency, and the moments that create downstream risk.
10 min read
Choosing a private deployment model for applied AI
Feb 05, 2025
A decision guide for VPC, on-premise, and air-gapped environments in regulated teams.
Monthly field notes
