BCG’s “Agentic AI Will Make the CMO’s Role More Consequential” argues that AI agents are becoming intermediaries in customer decisions. It asks CMOs to align brand promises with operational performance, turn customer intent into enterprise intelligence, and redesign marketing around journeys shared by humans and machines. The article is particularly convincing when it says that being visible to an agent is not the same as being credibly recommended by one.
The trouble begins with two unearned transfers: from what an agent knows to what a marketer can learn, and from what a CMO can see to what a CMO can change.
A customer may tell an assistant about dietary needs, budget pressures, and the real outcome behind a purchase. That is richer context than a search keyword. But it does not follow that this context becomes “a new class of customer intelligence” available to the brands being evaluated. Much of it belongs inside a private conversation mediated by a third-party platform. A brand may receive a referral or completed transaction while learning little about the deliberation that produced it.
That separation may be part of the value customers expect from an independent agent. An assistant working for the buyer should disclose only what is necessary to secure a useful result, not quietly turn intimate context into a marketing feed. If brands need platform access to recover those signals, the strategic advantage belongs partly to the platform controlling the interface. If they collect the same context through proprietary agents, customers may distrust advice from a system whose owner profits from the recommendation. BCG recognizes this conflict, but understates how much it limits the promised intelligence.
Even when an agent makes a brand’s failures more visible, the CMO usually does not own the machinery that caused them. Pricing integrity may sit with finance or revenue management. Availability belongs to supply chain. Product quality, fulfillment, claims handling, and service resolution have different executives, budgets, incentives, and systems. “Cross-enterprise governance” sounds like an answer, but shared meetings and metrics do not establish decision rights. AI can therefore make marketing more accountable for the complete customer experience without giving marketing the authority or capital to repair it.
This matters because a promise-performance gap is not principally a communications problem. If an agent finds repeated complaints about delayed delivery, the answer is not better structured content explaining the delivery policy. It is a better delivery operation. Making the evidence legible to machines may increase pressure to act, but the competitive advantage still comes from operational improvement. Agentic mediation changes the speed and scale at which failure is discovered; it does not change who must do the difficult work.
The recommendation to market to both humans and AI agents also risks treating machines as a stable new audience. There is no single machine decision maker. Different systems use different retrieval sources, commercial relationships, policies, prompts, memory, and ranking logic. Their answers change without the brand’s consent or even its knowledge. Metrics such as share of mentions, citations, and recommendation rates may help diagnose visibility, but they are proxies generated inside volatile environments. Optimizing them can recreate the worst habits of search marketing: chasing observable rankings while assuming they represent incremental demand or profitable growth.
BCG’s proposed operating-model redesign needs a harder layer beneath it. For every agent-mediated journey, companies should identify what customer context they can legitimately access, what evidence connects that context to revenue, who owns each operational defect, which executive has the right to reallocate resources, and how improvement will be measured against a baseline. Otherwise “customer intelligence” becomes information without access, and “brand stewardship” becomes responsibility without control.
The addendum is that agents may indeed make the CMO’s role more consequential, but not necessarily more powerful. They can expose the distance between promise and performance while widening the distance between customer insight and brand ownership. The winning company will not be the one whose marketing department monitors machines most aggressively. It will be the one that gives operational owners the evidence, incentives, and authority to fix what both customers and their agents can already see.