Cannes 2026 Called It the Agentic Era. The Product Isn't Shipped Yet.

Cannes Lions 2026 declared marketing's "agentic era" has arrived. It hasn't, not for you. Citi's own research team, fresh off the Croisette, called it "infancy." The vendors announced infrastructure. The agencies announced pricing pivots. Almost nobody announced a shipped, working, autonomous ad-buying system running real client budgets at scale. What is real and already changing outcomes is GEO, the discipline of getting your brand cited inside ChatGPT and AI Overviews instead of ranked on a results page. Owner-operators should ignore the agentic theater for now and start a GEO audit this quarter. That's the whole verdict. Here's the case for it.

I sat through a Cannes recap deck two weeks ago. A vendor rep walked a room of operators through "the agentic future of advertising" for forty-five minutes. Buyer agents. Seller agents. Agents negotiating with agents while marketers sleep. At the end, one owner in the back asked the only question that mattered: "Can I turn this on Monday for my $40,000-a-month ad budget?" The rep paused. "We're onboarding enterprise partners first." Translation: no. That pause told me everything about where this cycle actually stands.

What Cannes 2026 actually said, in plain terms

Citi Research sent a team led by internet analyst Ron Josey to Cannes this year. Their note, published on citigroup.com, is titled bluntly: advertising is entering the "agentic era." But read past the headline and the report is careful, almost cautious. It says adoption of agentic AI "remains in its early days." It says generative search is unlikely to "materially disrupt" incumbent Search or Social budgets in the near term. It says the biggest shift versus 2025 wasn't technology, it was tone: the industry stopped talking about AI as disruption and started talking about AI as enablement. Citi's own Cannes note is the softest declaration of a "new era" I've read in years, and that softness is the tell.

Contrast that with 2025. Last year at Cannes, the dominant mood was fear: AI was coming for creative jobs, for agency margins, for the whole model. This year the word changed from disruption to enablement, force multiplier, augmentation. That is not evidence the technology got dramatically better in twelve months. It is evidence the industry needed a new story once the disruption story stopped selling. I have watched this exact rhetorical pivot before, in dot-com pitch decks that went from "we will destroy retail" to "we help retailers," and in blockchain marketing decks that went from "we will kill the ad agency" to "we help agencies." The softening always arrives right when the product fails to show up on schedule.

Behind the scenes at Cannes, the honesty was blunter than the stage talk. Digiday's team spent the week collecting after-hours conversations, and the headline says it plainly: Cannes is selling AI transformation, but off stage the talk is about the bill. Executives described token costs running "vastly more costly than simply having engineers" and productivity gains landing closer to three times, not the ten times figures that get quoted from the stage. Digiday's Cannes briefing is worth reading in full if you want the unvarnished version of what I'm describing here.

The vendor announcements versus the shipped product

Here is the pattern I watched repeat across every Cannes recap I read. A platform announces "infrastructure" that "will allow" agents to transact. Note the tense. Amazon showed Alexa+ Agentic Ads. Fox announced an "end-to-end agentic advertising platform." Yahoo launched an open agent network. Nvidia demonstrated chips embedded across the ad stack. Zeta Global and Palantir signed a seven-year infrastructure deal. Every one of these is a real partnership. Almost none is a shipped product an owner-operator can log into today and run a live campaign through.

eMarketer's own coverage of the moment is candid about the timeline: PubMatic estimates 25% of ad buys will be executed autonomously by agents by 2028, rising past 50% by 2030. Those are the target dates, not this quarter. eMarketer's Cannes coverage calls this "moving from hype to actual deployment," which is generous. It is moving from hype to infrastructure contracts. Deployment for a small or mid-size operator is still years out, and hype cycles rarely arrive on schedule.

I have watched three of these cycles up close as an operator and advisor. The dot-com build-out promised every small business a storefront that would sell itself. Most small businesses spent two years and real capital on stores nobody visited. Blockchain-in-marketing promised loyalty programs and supply chains that would restructure entire categories, and almost none of it survived past the pilot. Metaverse marketing promised a new retail channel inside virtual worlds. Most of the activations built for it are abandoned digital real estate today. The pattern each time is the same: enterprise budgets fund the demo, trade press covers the demo, and by the time the tooling is stable enough for a founder-operator to touch without an engineering team, the narrative has already moved to the next thing. Agentic ad-buying is following that arc. It may still land. It is not landed.

Where the money is actually moving, right now, verifiably

The market sizing backs this up in an interesting way. Research and Markets put the AI-in-advertising market at $11.17 billion in 2025, growing to $36.34 billion by 2030, a roughly 26.7% compound annual growth rate. That's a real number, covered widely including by Yahoo Finance's coverage of the report. It is genuine growth. But notice what's inside that number: programmatic optimization, creative automation, predictive analytics, campaign intelligence tooling. Not autonomous agent-to-agent negotiation running unsupervised. The capital is flowing into AI-assisted advertising software that a human still operates. That is a very different asset than "an agent that buys your media while you sleep," and the difference matters if you are deciding where to spend your next dollar.

The part of this shift that is real, working, and billable today is GEO: generative engine optimization. This is the practice of structuring your content and brand signals so ChatGPT, Perplexity, and Google's AI Overviews cite you when a prospective customer asks a question instead of typing a keyword. Search Engine Land's 2026 guide lays out why this year is the tipping point: AI Overviews now reach over two billion monthly users, ChatGPT serves roughly 800 million weekly users, and traditional keyword search volume is measurably declining as buyers shift to asking assistants directly. Search Engine Land's GEO guide treats this as a discipline with its own audit process, its own KPIs, and its own tooling stack, because it is. Citi's Cannes note independently flags GEO as one of the few areas of genuinely growing budget commitment on the ground, not just on stage.

Here's my caution. GEO is becoming what SEO became fifteen years ago: a legitimate discipline that also spawned a large services-and-tooling market selling promises the underlying measurement can't fully back up yet. There is no mature, standardized way to measure "citation share of voice" the way Google Search Console measures rankings. Vendors are already selling audits, dashboards, and "22-point frameworks" against metrics that are, by their own admission, still being invented. GEO is the new SEO-industrial complex in the making. That does not mean skip it. It means treat every GEO vendor pitch with the same skepticism you'd apply to an SEO agency promising page-one rankings in thirty days.

The agency pricing tell

Watch what agencies do with their own business model, because it tells you more than what they say on a panel. The hourly rate, the unit agencies have sold for decades, breaks down completely once AI compresses a deliverable timeline three or four times over. An agency still billing by the hour on AI-assisted work watches its revenue fall as its output quality rises. That is an unworkable position, and the smartest shops know it. WPP, the largest holding company in the industry, now derives 20 to 25% of net sales from performance-linked fees and has publicly committed to moving away from time-and-materials billing altogether. A 2025 survey of more than 180 agencies found roughly a third had already fielded client requests for an "AI discount," with about half expecting one soon. Digital Applied's 2026 agency pricing guide lays the mechanics out in detail.

Read that pricing shift as a confession. Agencies are hedging against a future where AI does more of the work and clients know it. Outcome-based pricing is the agency version of "we're not sure what this technology is worth yet, so let's tie our fee to what it produces." That is a sound instinct, and a sign the industry has not settled on what agentic tools are worth. It should give any owner-operator pause before overpaying for a vendor's confidence.

The Owner-Operator Frame

I run every new marketing trend through the same filter before I let a client near it. I call it the Owner-Operator Frame: does this change what you actually do Monday morning, with the budget and headcount you actually have, or is this a story built for an enterprise CMO with a nine-figure budget and a data science team? Most of what got announced at Cannes this year fails that test immediately. Agent-to-agent programmatic infrastructure is being built for holding companies and platforms with the engineering bench to integrate it. You do not have that bench. You are not the audience for that announcement, no matter how many trade press headlines use the word "you" in the subhead.

GEO passes the Owner-Operator Frame, at least partially. You can audit your own AI citation visibility this week with a handful of manual prompts across ChatGPT, Perplexity, and Google's AI mode. You can restructure your best three pages with clear headings, direct-answer paragraphs, and FAQ blocks without hiring anyone. That is a Monday-morning action. What fails the frame inside GEO is the $5,000-a-month "AI visibility platform" retainer sold on metrics nobody can fully audit yet. Take the discipline. Skip the premium tooling until the measurement matures.

What I'd tell a client this week

Compartmentalize the two conversations. Agentic ad-buying is a casualty drill you watch from a safe distance, not a system you build your Q3 plan around. Log the vendor names, note who ships real self-serve product versus press releases, and revisit in twelve months. Do not sign anything requiring you to be an "early enterprise partner." That phrase means you are the unpaid QA department.

GEO is different. It is an asset you can build now with existing content and existing headcount, and it compounds the same way good SEO always compounded: slowly, then durably. Audit where your brand shows up, or doesn't, in AI-generated answers for your top ten buyer questions. Fix the structural basics: clear headings, direct answers up front, FAQ schema, an updated "last modified" date on your cornerstone pages. That's real, sellable groundwork toward a brand asset that a future buyer of your business will actually value, unlike a vendor subscription you can cancel anytime with zero residual worth on your balance sheet.

Doctrine Connection

Doctrine Connection: Verification beats optimism

Every hype cycle sells itself on optimism: believe first, verify later, don't get left behind. The Owner-Operator Frame exists to reverse that order. Verify what a technology actually does, on your budget, on your timeline, before you let anyone's optimism become your operating plan. Citi's own analysts hedged their "agentic era" headline with the word "infancy" in the same report. If the bank writing the bullish note won't remove its own hedges, you shouldn't remove yours either.

Frequently Asked Questions

Jeff Barnes, MBA has no personal position in any company, fund, or platform named in this article. demg.ai has no current commercial relationship with any party mentioned. demg.ai provides marketing strategy and education for owner-operators, not investment advice.