How Enterprises Build AI That Lasts.
Models are increasingly shared. The competitive ground has moved to what an organization builds around them — a context layer of its own, and the discipline to tell proven value from a working demo. Both are paid for in operating-model change.
Advance Praise
"Generative AI models are undoubtedly amazing, but this refreshingly honest book puts them in their place. Unless your organization redesigns work around AI capabilities, captures organizational context, upskills your people, and puts the AI system into full production, you won't get economic value. It's all hard work, but the sooner you get started the sooner you'll succeed with AI."
"In Power and Prediction, we argued that the real value of AI emerges not just from better point solutions, but from redesigning systems that properly utilize machine prediction and human judgment. Operating Intelligence is an unusually practical guide to that system-level work, showing executives how to move beyond demos toward governed context, proof of value, and operating-model change that creates and compounds value. For leaders seeking to turn clever neural nets into durable enterprise advantage, Brandon Chiazza has written the operating manual."
"Cheap prediction leads to the question: which decisions do you redesign? Operating Intelligence provides answers based on Chiazza's real-world experience in AI strategy."
"A proof of concept shows that the technology can perform.
A proof of value shows that the work improves."
— Law 3
The book
Full Edition
Available now
$12.99 Kindle
$29.99 paperback · 477 pages
All 477 pages. Strategic foundation, data and governance, system-building (LLMs, prompts, RAG, agents, multi-agent), operating chapters (tooling, security, cost), value, adoption, future. The reference manual for the people who will build the program and then have to run it.
Buy on Amazon →Sample Chapter
Free
PDF · Chapter 6 · no signup
Chapter 6, Data Governance for AI, as it appears in the book: the MyCity, MyAI case, the AI Data Trust Model, and the controls that separate deploying AI from governing it. Figures and references included. Share it inside your organization.
Read Chapter 6 (PDF) →Kindle and paperback are both on Amazon. If the button does not find them yet, the listing is still propagating; join the list and we will send the direct links. Corporate and classroom orders: hello@modali.com.
The framework
The argument of the book, compressed into twelve operating principles. Short enough to quote in a board meeting. Substantive enough to hold the program accountable to them.
Models are the smallest part of enterprise AI advantage.
The context layer is the part of the AI stack that does not depreciate when the model changes.
A proof of concept shows the tech can perform. A proof of value shows the work improves.
Baseline before you build. A weak baseline is more useful than no baseline.
Hallucinations are a structural property of next-token prediction, not a bug.
A disclaimer is not a control.
Bounded autonomy: permissions, reversibility, and escalation are first-class design decisions.
The evaluation harness is the precondition for everything else.
Three words for value: projected, realized, banked.
Every AI system needs a written rule under which it will be taken offline.
Authority should match accuracy.
Operating-model change is not the consequence of AI adoption. It is the price.
Working artifacts
Three downloads adapted from real engagements.
Self-Assessment
Twelve questions, one per Law, scored 1–5, with a maturity readout in four bands. Score it yourself, or point it at a single initiative. The most useful version has each function score the same program, and the disagreements are the finding.
Download (with newsletter signup) →Charter
The operating contract between a proof-of-value team and the executive sponsor: baseline, value hypothesis, target users, evaluation method, eval owner, adoption criteria, risk controls, scale-or-stop rule. Fillable template plus the two Framework Quality Tests.
Download (with newsletter signup) →Diagnostic
Twenty questions across seven sections assessing whether the organization has a named, governed context layer. Maturity readout in four bands; the section minimum matters more than the total.
Download (with newsletter signup) →Sample chapter
Chapter 6 — Data Governance for AI — opens with MyCity, MyAI, the public-sector chatbot that operated for 28 months under official city branding while answering legally consequential questions wrong. The chapter develops the AI Data Trust Model and the structural controls that distinguish "we deployed AI" from "we govern AI." Free, no signup, share it with your team.
Read Chapter 6 (free PDF) →
Author
CEO of Modali Consulting, where his work focuses on helping enterprises adopt AI in ways that compound value rather than capability.
Faculty at the Cornell Brooks School of Public Policy. Formerly Chief Technology Officer at the New York City Mayor's Office of Contract Services, where he led major digital-transformation work in procurement and public-sector technology.
Named co-inventor on two U.S. patents in applied AI. Authored the MyCity, MyAI public-policy case study at Cornell — the public-sector AI deployment that frames the governance argument across Chapters 6, 16, and 19 of the book.
Speaking · Workshops · Advisory
The Twelve Laws Self-Assessment runs as a 90-minute leadership workshop. The Proof-of-Value Framework and the Context Layer Diagnostic anchor longer engagements. Modali Consulting works with enterprises on the operating disciplines this book develops — context-layer architecture, proof-of-value design, AI governance, and the operating-model change that turns AI capability into banked value.
For instructors and program directors
Operating Intelligence is built for graduate courses and executive certificate programs in AI strategy, AI governance, digital transformation, and public-sector technology. Eighteen chapters, the Twelve Laws, a public-record central case (New York City's MyCity chatbot), and three working artifacts students can score and fill in. The teaching guide below has chapter summaries, learning objectives, discussion questions, and applied assignments for every chapter, along with three course maps: a twelve-week graduate seminar, a six-session executive module, and a three-hour workshop.
Request a paperback examination copy for a course or program you direct. Tell us the course, the term, and the expected enrollment, and we will mail one at no cost. Bookstores order through standard wholesale channels; for 25 or more copies, ask about direct pricing.