The first edition of The Agentic CMO shipped in June 2025. In the twelve months since, the technology stack underneath every framework in the book has been rebuilt at least once. The EU AI Act has crossed two enforcement gates. NVIDIA has shipped an enterprise-grade agent operating system. OpenAI and Google have published commerce protocols for AI agents to transact with brands. The Spencer Stuart and Gartner numbers on CMO headcount, AI literacy, and adoption maturity have moved enough to require a full statistical refresh.
A second edition was not a marketing exercise. It was an obligation to readers buying the book at the wrong end of a compression cycle.
This article is the orientation. What the second edition adds, what stays the same, and what every CMO should be doing in the next ninety days regardless of which edition they happen to be reading.
The Numbers That Forced a Rewrite
McKinsey’s modelling puts the annual value of generative AI at up to $4.4 trillion and the broader AI opportunity at up to $13 trillion, with marketing and sales as the single largest functional pool inside both. McKinsey’s own breakdown estimates AI could drive around $463 billion in marketing productivity gains, with agentic systems accounting for more than 60 per cent of the new value. The opportunity is large. The execution gap is what should keep CMOs awake.
The latest data from Gartner says it cleanly. Seventeen per cent of organisations have deployed AI agents to date. Sixty per cent expect to within two years. Only one in five has a mature governance model for autonomous agents. Translation: four out of five organisations deploying autonomous systems are doing so without the governance infrastructure to manage them safely at scale.
Adobe Analytics reported a 4,700 per cent year-on-year increase in US retail traffic from generative-AI browsers and chat services in July 2025. The base is small. The slope is what matters. Adoption curves do not climb in straight lines, and the early steepness of this one is doing what early steepness usually does: setting the conditions for the inflection that follows.
These are not predictions. They are measurements. The first edition assumed the inflection was coming. The second edition is written for the readers who realise it arrived.
The Bet Underneath Both Editions
One argument runs through every chapter, and it is the single most contrarian position in the book.
Most agentic marketing pilots will not fail on governance. They will fail on brand coherence across a fleet of agents that do not share a memory, a synchronised tone, or a living style system.
Governance is what regulators force. Brand coherence is what decides whether a portfolio of agents earns its capital or dilutes it. The two are not the same. A CMO who lets an agent fleet operate without a coherence-allocation discipline will, month by month, produce output that is locally on-brief and globally off-brand. Each agent does what its individual brief asks. The fleet as a whole pulls the brand toward whatever cluster the underlying foundation model finds most natural — a cluster chosen by training data the CMO did not select.
The second edition is structured around that bet. Every chapter sits in service of it. If the bet is wrong, the book is still useful. If the bet is right, the apparatus the book provides is what separates the CMOs who compound from the CMOs who watch their brand drift.
What Is New in the Second Edition
Five substantial additions. None is cosmetic.
Chapter 6 — the portfolio-manager framework — is now set out in operating-manual detail. Position sizing with an actual formula. The seventeen-per-cent cash reserve as a non-negotiable line item, not a footnote. Exposure budgeting across brand, legal, customer, operational, vendor, and model risk. Lifecycle gates: incubation, production, retirement, with explicit retirement triggers. The first edition sketched these. The second edition turns them into a fund manager’s playbook.
Chapter 8 absorbs the deployer-side liability case under the EU AI Act. The Act’s full requirements for high-risk systems take effect on 2 August 2026, with fines up to €15 million or three per cent of global turnover. The standard SaaS contract was not drafted for autonomous systems that publish sixty thousand pieces of customer-facing content a year. Appendix B now carries a complete model-clause kit — ownership perimeters, training rights, decision-quality SLAs, drift SLAs, explainability SLAs, indemnities for autonomous failure modes, termination as a planned capability — with three-position drafting for every clause.
Chapter 14 treats the customer who is no longer a person. OpenAI’s Agentic Commerce Protocol and Google’s Agent Payments Protocol are both shipping. McKinsey projects three to five trillion dollars of agentic commerce globally by 2030. Generative Engine Optimisation is now a discipline, not a slide. The dual-track marketer — fluent in human persuasion and machine optimisation — is the role most enterprises do not yet realise they need.
Chapter 5 — the organisational chapter — now incorporates a substantive treatment of European works-council and codetermination obligations. The variable that turns an eight-week vendor Gantt chart into a seven-month deployment in Germany, France, Italy, and the Netherlands. Most US-headquartered firms are still pricing this wrong.
Every quoted statistic has been refreshed against 2026 sources. Anthropic’s March 2026 Labour Market Impact Report. Spencer Stuart’s 2026 CMO survey. Gartner’s AI Blind Spot research. The Adobe traffic figure. Refreshing the data was the last stage of writing the second edition and the one that most clearly justified doing it.
The Method Is the Message
A confession sits in the introduction of both editions, but it carries more weight in the second.
This book was written using the methods it teaches. A research agent synthesised market data and identified pattern connections across industries. An editorial agent maintained narrative coherence across complex frameworks. Neither could have done the work alone. Neither could the human author have done it without them.
That is not a marketing flourish. It is the demonstration that this book exists to teach. Hybrid intelligence is not a slide. It is an operating model. When it is run honestly — bidirectional learning, failure-mode complementarity, explicit arbitration — it produces output neither side could have produced separately. When it is run as a press release, it produces a manuscript that sounds like every other manuscript on AI written by every other consultant in 2026.
The reader is invited to apply the same test to anything written under the label of AI strategy. If the author cannot explain which parts of the work the machine actually did, the author is selling a posture, not a method.
The Five Parts, Reread for the Second Edition
Part I makes the strategic case. Why agentic AI is a structural shift, not a tooling upgrade. The Autonomy Spectrum that separates rule-based automation from predictive AI from generative AI from agentic systems. The CMO’s evolving mandate from brand guardian to intelligence orchestrator to portfolio manager.
Part II designs the hybrid organisation. The capital case for agentic investment. The architecture of human-AI collaboration — hub-and-spoke, embedded, hybrid collaborative. The cultural and talent shifts that determine whether the architecture survives contact with reality. The portfolio framework that sizes positions, budgets exposures, and retires zombies.
Part III is implementation excellence. The pilot-to-scale journey, with the crawl-walk-run methodology that compresses to weeks what most enterprises stretch to quarters. The deployer-side liability case under the EU AI Act. The technology stack — perception, intelligence, orchestration, execution, learning — that production-grade agentic marketing requires.
Part IV leads into the future. Measurement when causality breaks. Ethical leadership when dark patterns emerge from gradient descent without human intent. Customer experience design that survives the personalisation paradox. Competitive strategy when Klarna can cut marketing-supplier spend by twenty-five per cent and AI-native challengers attack with teams of dozens.
Part V is the playbook. The customer who is an algorithm. The talent restructure that produces fewer, more senior marketers. Industry-specific applications where sector gravity reshapes implementation more than technology differences do. The CMO’s three evolutionary paths — Expand, Go Deep, Go Agentic. The ninety-day transformation plan where day one decisions determine decade-long value.
Eighteen chapters. One spine. The portfolio manager mandate, instrumented with the apparatus required to discharge it.
What This Book Cannot Give You
Conviction. That is yours.
Every framework is built on evidence. Every case study is real. Every regulatory citation is enforceable law. Frameworks do not transform organisations. Leaders do.
The shift from campaign manager to intelligence orchestrator requires a CMO willing to stake professional credibility on an operating model that most peers have not yet adopted. That is not a comfortable position. It never is when you are at the front of the column.
The risks of inaction are real. Compressed transformation windows. Exponential competitor advantages. The erosion of relevance. The risks of action are equally real. Organisational resistance. Pilot failures. Budget scrutiny. The career exposure that comes with championing something new.
Both sets of risks are real. Only one set gets worse with time.
The Decision the Second Edition Forces
By the end of 2027, two adjacent CMO archetypes will be visible at most enterprises. The first runs a marketing function whose headcount has fallen twenty-five to forty per cent and whose remaining people are visibly more senior, better paid, and shipping more than the larger team they replaced. The second runs a marketing function the same size as today, with the same titles, the same seniority distribution, and a flat or declining contribution to revenue.
The board will ask each of them the same question. One will have an answer the board recognises. The other will not.
The first edition was written for the CMOs who suspected the answer existed and wanted the frameworks. The second edition is written for the CMOs who have run out of room to defer the choice.
Eighteen weekly articles will follow this one. Each takes the spine of a chapter and turns it into something usable in the next quarterly review. The series closes with a synthesis and a direct invitation to the twenty-seven live prompts that turn every framework in the book into a working artefact a CMO can put in front of a board on Monday.
Start with the chapter that maps to the conversation already on your desk. Read in any order. The frameworks compound regardless of sequence.
The window for catching up closes faster than most boards realise. The CMOs already moving are not waiting for certainty. They are building the apparatus that compounds while their peers debate the timing.
The question is which camp you join this year, not next.
Keep Reading
That’s all for this week book chapter summary, come back next Monday for the next chapter summary.
The Agentic CMO - Second Edition is available today in hardcover, paperback and ebook.
Disclaimer: The views and opinions expressed in The Agentic CMO, Chronicles of Change and on my social media accounts are my own and do not necessarily reflect the official policy or position of S&P Global.