At Cannes Lions, where the industry is predisposed to turn every product announcement into a revolution, Digitas CEO Amy Lanzi made a deliberately unglamorous point: AI won’t save advertising.
That is less of a contrarian soundbite than a diagnosis. The problem in advertising has rarely been a lack of models. It has been the inability to turn marketing into something that behaves like a governed system: a repeatable pipeline from data collection to analysis, activation, measurement, and iteration. Lanzi’s argument, as reported by The Verge, was that business results still depend on data and analytics first. AI can improve the machinery, but it does not replace the machinery.
That distinction matters because the current phase of AI adoption in marketing has often been framed as if the technology itself could repair weak measurement, fragmented data, and inconsistent operating practices. Digitas is pushing in the opposite direction. Rather than treating AI as a layer of novelty on top of existing workflows, the agency is reorganizing around what it calls AI-driven intelligence — a structure that implies process discipline, governance, and system design.
The most visible sign is the new leadership model Digitas is putting in place. The reported roles — Chief Intelligence Officer, Chief Systems Officer, and Chief Transformation Officer — are telling precisely because they do not sound like temporary innovation theater. They map to three problems the industry has struggled to solve for years: how to make intelligence legible to the business, how to make the underlying systems reliable enough to trust, and how to drive change through organizations that are usually optimized for campaign velocity rather than operational coherence.
That reorg is a useful signal for anyone building or buying AI-enabled marketing tooling. It suggests the center of gravity is shifting away from feature-level AI and toward the architecture needed to support it. In practice, that means cleaner data pipelines, stricter governance, more consistent metric definitions, and a better understanding of where automated decisions are allowed to happen. If a company cannot unify identity, conversion, and performance data with enough rigor to support decisioning, then even strong models will only produce faster confusion.
This is where the phrase “scalable, governed system” becomes more than consulting language. Advertising organizations have long depended on a patchwork of platforms, point solutions, agency workflows, and manual reconciliation. AI does not remove that complexity; it amplifies the cost of bad foundations. A model trained on incomplete or inconsistent data may generate plausible outputs, but those outputs will not reliably improve return on ad spend, creative strategy, or budget allocation unless the data architecture underneath is coherent.
Digitas’ move also hints at a broader reset in how agencies position themselves. The easy pitch is that AI makes every vendor smarter. The harder, more credible pitch is that value comes from integrating AI into operating systems that can actually be measured. That changes what clients should ask for. Instead of demanding more AI features in isolation, they will likely care more about whether a platform can standardize measurement, expose data lineage, support human review, and connect experimentation to financial outcomes.
For ad-tech vendors, that is a meaningful shift in buying criteria. The product conversation moves from “what can your model generate?” to “how does your stack govern outputs, preserve auditability, and fit into an existing analytics runtime?” The winners in that environment are less likely to be the loudest AI demos and more likely to be the platforms that can prove they fit inside a disciplined marketing system. Standardized metrics, interoperable data models, and workflow controls become strategic features, not back-office concerns.
The Cannes context matters here as well. July 2026 coverage from the festival keeps surfacing the same underlying message: the industry is still captivated by AI, but the most credible operators are talking about infrastructure, not spectacle. Digitas’ public framing is notable because it treats intelligence as an organizational capability, not just a vendor category. That is a more demanding view of change, but also a more durable one.
For product teams and operators, the watchlist for 2026 and 2027 is straightforward. First, look for evidence that data maturity is improving: better identity resolution, cleaner event pipelines, more consistent attribution logic, and stronger governance over inputs and outputs. Second, look for AI workflows that are embedded in decision systems rather than bolted onto campaign tools. Third, pay attention to whether organizations can tie experimentation to ROI in a repeatable way, not just in pilot projects.
The real lesson in Lanzi’s comments is not that AI is overrated. It is that AI is now mature enough to expose whether a marketing organization has built the systems needed to use it well. That may be a less exciting thesis than the industry wants, but it is a far more actionable one.
