Archive
Notes on agentic analytics, leadership, systems, and what it actually takes to ship.
Adoption is easy to count. Compounding is what matters. The three-pillar framework I use for AI transformation - and why investing in just one pillar produces a pilot that plateaus.
Five public-model calls. Five outcomes that broke them. A weekend stress-test of the modern NBA analytics stack against the Knicks' championship run, and what it reveals about the gap between what models measure and what actually decides a series.
AI didn't kill data science. It killed the version of data science that was just running queries and writing decks. The work moves up the stack — from analysis to system, from model to harness.
ML platform discipline transfers to AI analytics systems — feature stores become semantic layers, drift detection still applies. The serving layer is where the bar goes up, and eval is the piece without a shared standard.
Anyone can build now. The next decade belongs to whoever closes the last 20% with real memory, context, and precision.
The labor market isn't disappearing. It's repricing around who can think, decide, and own outcomes.
Most teams still define agents as human roles. The longer-term shift is toward workflow-native systems, review discipline, and context infrastructure.
What actually shortens the gap between insight and action? Decision latency is the metric that matters.
There's a difference between using AI and restructuring around it.
Capability is no longer the bottleneck. Operational economics, context engineering, and governance are.
What separates people who thrive in AI-native orgs from those who are stuck.
Decision latency is the real metric, not chat interfaces or dashboards.
Most companies don't have an AI problem. They have a data readiness and decision problem.