Book Series · 2024–2026
Four books on designing, building, and governing agentic AI products, and on keeping the human judgment they route through.
These books were written from a practitioner’s seat, not an analyst’s. Each one is a field guide to a specific problem: what the PM job actually requires, what the team has to hold, and what happens when the agent runs without a design behind it.
They were written in order, but each stands alone. If you are new to agentic product management, start with Book 1. If you are building something specific, start where the problem is. If you have been operating agents long enough to feel the supervision skill quietly erode, start with Book 4.
The full PM lifecycle for agentic products: deciding whether to build, designing behavior and evals, measuring drift, and managing the human system around the agent. Thirteen chapters.
Every agentic product is two products. Most teams build the first and assume the second takes care of itself. It does not. Twenty-two chapters on the supervisory layer’s four dimensions, failure modes, and how to build it deliberately.
The prior books were written from the PM’s seat. This one zooms out. The unit of failure is the seam between roles, not the capability of any one of them. Twenty-six chapters.
The first three books rest on one assumption nobody audited: that the practitioner in the seat still has the judgment the structure routes through them. This book is the deliberate-practice curriculum that keeps that assumption true. Fifteen chapters and a full appendix set.
Seven skills covering the agentic PM lifecycle, distilled from the four books. Install in Claude Cowork and invoke at the moment of need. Stage 1 of three.
Yoram Friedman is a physician and senior product manager who has spent fifteen years building enterprise cloud, data and platforms at SAP, Walmart, and Microsoft. He trained as a physician in anesthesiology and radiology and writes about healthcare AI, agentic systems, and what it actually takes for AI to move from pilots to production in regulated environments.
The regulatory and design default for AI is “keep a human in the loop.” Sustained AI use erodes the human’s ability to perform that loop’s function. This series is for the people who have to design supervision that does not assume the supervisor is the unimpaired version of themselves they were on day one.