Madison Avenue meets the algorithm: what the Cannes boardrooms said out loud

The Cannes Lions International Festival of Creativity is, on its surface, a celebration of advertising craft. Underneath, it has become an annual checkpoint where the world's largest marketing platforms and their biggest clients compare notes on a much harder question: who actually owns the consumer relationship when the buying and the selling both happen inside an algorithm. On 3 July 2026, Variety's "Strictly Business" podcast convened leaders from NBCUniversal, IBM, State Farm, Autodesk and Coinbase for a 30-minute session that read less like a panel and more like an industry-wide status report on the post-cookie, post-attribution, AI-saturated marketplace.

What emerged from the conversation is a thesis the major trade press has been edging toward for two years: the marketing stack is being rebuilt around AI, and the companies that control the underlying data infrastructure — not the creative agencies or even the media sellers — will collect the largest share of the new economics. It is a structural shift, and the participants were unusually direct about it.

The new gatekeepers are not the networks

For most of the television era, Madison Avenue answered to a small number of large media companies. NBCUniversal sat comfortably inside that club. On the podcast, the company's leadership was candid that the relationship between content owners and marketers has been unsettled by the same forces that upended every other consumer-facing industry: on-demand viewing, identity fragmentation, and the slow disappearance of the third-party tracking cookies that once allowed a brand to follow a single household across screens.

The replacement is not a single product but a portfolio of "identity," "clean-room," and "AI-curated audience" offerings — joint data environments where a brand can match its customer file against a media partner's first-party data without either side ever seeing the other's raw records. NBCUniversal's contribution to this conversation, as reported by Variety, is that it now treats its owned-and-operated streaming inventory less as a media business and more as a data business wrapped around programming. The implication for advertisers is that the negotiating counterparty is no longer the network sales executive; it is the data partnership team, which has very different incentives and very different margins.

AI as creative, AI as middleman

If the data layer is the new gatekeeper, AI is the new middleman — and the panel split roughly along a fault line that has nothing to do with industry and everything to do with business model. Coinbase, the crypto exchange now rebuilding itself around regulated stablecoin rails and a US-compliant trading product, framed AI as a customer-acquisition tool: a way to compress the cost of converting a curious browser into a verified, funded account. For a business that has historically spent heavily on performance marketing against a thin-margin product, every basis point of conversion efficiency matters disproportionately.

IBM and Autodesk, by contrast, are selling to enterprise buyers. Their interest in AI is upstream of the consumer entirely — AI assistants inside the product that reduce churn, AI co-pilots inside the design workflow that justify a higher seat price, AI agents that summarise a client's own data inside a dashboard. State Farm sits between the two: a century-old mutual insurance franchise using AI to triage claims and personalise quotes while still paying for Super Bowl-scale reach to defend a brand that is, by the company's own admission on the podcast, only as valuable as the trust it carries.

What unites these otherwise incompatible businesses is a shared assumption: the next decade of marketing efficiency will be measured by how well a company can embed AI into both the buying side (media planning, attribution, creative iteration) and the selling side (service, personalisation, retention). The agency of record, in this framing, becomes a systems integrator rather than a creative auteur.

The structural read

Strip the industry jargon away and what the Cannes conversation actually documents is a familiar pattern in a new industry. The platform layer of digital advertising — Google's search and display inventory, Meta's social graph, Amazon's purchase-intent signal, and increasingly a handful of AI-native intermediaries — captures an outsize share of every marketing dollar because it sits at the point of friction between intent and fulfilment. Creative agencies, even the holding-company giants, sit one layer back from that point and are therefore one layer back from the margin.

The diversification strategies outlined on the Variety podcast — NBCUniversal's identity partnerships, IBM's enterprise AI services, Coinbase's regulated stablecoin distribution, Autodesk's AI-augmented design tools, State Farm's claims automation — all push, in their different ways, toward owning a piece of the data or AI layer directly. None of the executives used the phrase "vertical integration," and none needed to. The implication is that the next round of marketing-industry consolidation will not be the holding companies buying more agencies. It will be the brands and the platforms buying or building the AI and data infrastructure they currently rent.

That is also where the most obvious counter-narrative sits. The consumer-protection and competition-policy argument — voiced across the European Commission, the UK Competition and Markets Authority, and US state attorneys general — is that concentrating the AI and identity layer inside a small number of vertically integrated players produces the same privacy and market-power harms the cookie era produced, only faster and with less auditability. The companies on the Variety panel would, privately, agree; their public answer is that first-party data partnerships and clean rooms are a sufficient governance response. Whether that answer satisfies regulators over the next 18 months is one of the more consequential open questions in the industry.

Stakes and what to watch

The near-term stakes are concrete. Brands that cannot — or will not — invest in their own first-party data infrastructure will pay a steadily rising tax to the platforms that have. Mid-tier agencies that do not reposition as AI-systems integrators will see their share of the marketing dollar compress further. Independent media owners that cannot offer a clean-room identity product will be relegated to remnant inventory in the planning tools that AI agents will use to assemble campaigns by default.

Over a longer horizon, the more interesting question is whether the AI-driven marketing stack produces a measurably better outcome for consumers, or merely a more efficient extraction of attention and wallet share. The executives in Cannes argued, with varying degrees of candour, that personalisation at this scale is fundamentally a service improvement. Regulators, civil-society groups, and a meaningful slice of the academic literature argue that it is a surveillance improvement dressed up as service. The truth, as usual, will probably settle somewhere inconveniently between the two.

What remains genuinely uncertain is the speed of the transition. The companies on the Variety podcast have the capital, the regulatory access, and the engineering talent to execute the strategy they outlined. The open question — which the panel did not pretend to answer — is whether the broader industry, and the consumers inside it, will have any practical choice about following them.

— A Monexus desk note. Wire coverage of Cannes Lions tends to read as either trade-press stenography or celebrity gossip. We treated this one as an industry status report: the actual news is the convergence of five otherwise incompatible businesses on a single structural bet about where the marketing margin moves next.