Notes
Field notes from shipped work.
Short essays on agentic AI, consulting, and what actually works in the room with operators.
- ·4 min readroifinance
Cost-per-run is the only number your CFO actually wants
Most agentic-AI pilots skip the unit economics. Here's the napkin math that decides whether your project pays back in ten days or leaks for a year.
- ·4 min readagentsstrategy
Why the agent your enterprise actually needs is boring
Procurement keeps getting pitched flashy multi-agent demos. The agent that ships value is a single tool-using loop on a workflow you already have a runbook for.
- ·5 min readagentsevaluation
The five-day eval rubric: how to grade an agent before you ship it
Teams that ship agents rarely write the rubric first, so they can't tell when the agent regresses or defend cost-per-run to finance. A five-day exercise that fixes both.
- ·1 min readstrategyarchitecture
Build or buy: the three cuts that actually matter
Most build-vs-buy frameworks are too abstract to use. Three questions that force a real answer in under an hour.
- ·1 min readconsultingroi
The ten-day payback rule
If your first AI project can't pay for itself in ten days, you picked the wrong project. The shape of engagements that actually ship.
- ·1 min readagentsstrategy
Agentic AI is not a chatbot
If your AI still needs someone to prompt it, it's not agentic. Here's the mental model that actually matters, and how to tell when you need it.