Loading...
Marc Pritchard's modular agency framing, the Klue credential breach, and an 85% vote for outcomes-based pay all point at the same unoccupied strategy layer.
The most powerful CMO in advertising stood on the Palais stage this week and said out loud what the agency holding companies have been trying not to hear. Marc Pritchard, Procter & Gamble's chief brand officer, told the Cannes Lions audience that brand building is now a "constant sprint" and that an "AI-plus-human approach" is "rocket fuel" — language he paired with a named endorsement of a "modular agency model" that breaks the holdco mold. The same week, a competitive intelligence vendor confirmed that a credential issued in 2022 — four years old, never decommissioned — was the entry point for hackers who exfiltrated names, addresses, support cases, and CRM records from LastPass, BeyondTrust, Huntress, Recorded Future, Tanium, Jamf, and Salesforce itself. An independent agency launched an "open orchestration layer" explicitly aimed at the holdcos' proprietary AI stacks. And 85% of agency leaders told a Cannes-timed survey that outcomes-based payment — not hourly fees — is the future of agency compensation.
None of those four signals are about Autostrat. All of them point to the same gap.
For three years we have been arguing that the strategy tier of the agency market is structurally unoccupied by both the holding companies and the research tools. The holding companies are now busy building AI stacks that route work through their proprietary agents. The research tools sell access to data and leave the synthesis to the buyer's team. The buyer is left holding an orchestration problem that no single vendor is accountable for. This week's Cannes news made that problem visible, named, and endorsed by the world's largest advertiser.
P&G is not a minor brand. It is the single most credible buyer-side voice in the advertising industry, and Pritchard's framing of a "modular agency model" is the first public repudiation of the holdco-orchestration thesis from a buyer with the scale to make the repudiation stick. According to Marketing Week, Pritchard described the industry as entering a phase of creativity "unlike anything" previously experienced, with AI amplifying human judgment rather than replacing it. The phrase "rocket fuel" was deliberate. AI is the propellant; human accountability is the craft.
The buyer's modular frame is not a defense of the holdco stack. It is a defense of the buyer's right to pick the best agent or capability per workstream — and to keep the named accountable owner of the strategy on the buyer side, not on the agency side. That is a structurally different model from the one WPP, Publicis, and Omnicom have been building since January, and it is the first time the buyer's preferred answer has been articulated in public by someone with P&G's reach.
The same week, Bob Lord, CEO of Horizon Global, took out a sponsored essay in The Drum arguing that "the entire agency model" has become opaque and that "clients and their needs are getting lost in the mix." Horizon is one of the largest independent agencies in the world. When the agency's CEO and the world's largest advertiser arrive at the same conclusion in the same 72 hours, the conclusion is no longer fringe.
The Klue breach, as confirmed by TechCrunch, has moved past the "incident" phase and into the "chain-of-custody failure" phase. The credential was issued in 2022 for a limited pilot with a third party. Klue had years to decommission it. It did not. The credential then sat on a four-year-old OAuth integration with live access to customer Salesforce instances at security vendors whose entire value proposition depends on operational security discipline.
This is not a story about a vendor that got hacked. It is a story about a category — competitive intelligence as a software subscription — that structurally requires buyers to put their strategy data into a third party's database, where it is subject to the operational hygiene of that third party and the credential-management discipline of every integration the third party has ever shipped. The CI category has spent five years telling CMOs that the right answer to strategy data management is "more dashboards, more access, more agents." The Klue story is the second-order proof that access and exposure scale together.
LastPass confirmed to affected customers that names, phone numbers, email addresses, physical addresses, support case records, and sales-related data were all exfiltrated. Salesforce and Gong have permanently disabled the Klue integration. The chain-of-custody failure spans at least twelve named security vendors and is ongoing. For any CMO whose strategy data lives in a third-party CI subscription, the math of "do I keep my intelligence in someone else's database" has changed.
MediaPost's coverage of the MediaSense Cannes report puts a number on the consensus: 85% of agency leaders say outcomes-based remuneration is the preferred long-term compensation model. The FTE-based billable hour — the operating model of every major holdco for forty years — is now formally under siege in the trade press.
This is the third signal. The buyer's preferred payment model, the buyer's preferred accountability model, and the buyer's preferred modular structure are all publicly aligned against the holdco stack in the same week. The category definition for "what the buyer actually wants from an agency in 2026" has been written in three places by three different constituencies in the same seventy-two hours.
The piece that is missing from all three signals is the same piece. Who owns the named, accountable, appeal-ready interpretation of the agentic output? Not the agent. Not the holdco platform. Not the CI tool with the four-year-old credential. The buyer. And the buyer's in-house team is structurally not staffed to do this work — that is why the buyer hired an agency in the first place. The buyer's modular frame requires a new kind of partner: one that delivers outcomes, takes named accountability for the recommendation, and is paid on the result rather than the hour.
The buyer-side answer is not "build it in-house." The buyer's modular frame is a repudiation of the holdco stack, not a repudiation of agency relationships. The buyer's preferred answer is a named accountable partner who operates inside the modular architecture, owns the interpretation of what the agents produce, and answers to the buyer — not to a holding company platform.
Three implications for CMOs building their 2027 agency roster:
First, ask any prospective partner who is named on the deliverable. Not "who is the team" — who is the individual whose name is on the recommendation, who can be appealed to, and who is accountable if the recommendation turns out to be wrong. The holdco stack cannot answer this question. The CI tool cannot answer this question. A named strategy partner can.
Second, ask any prospective partner where the strategy data lives. If the answer is "in our platform" or "in a third-party database we integrate with," the buyer is taking on chain-of-custody risk that compounds with every quarter of subscription tenure. If the answer is "we deliver the decision; the data lives where you put it," the buyer is buying outcomes, not exposure.
Third, ask any prospective partner how they are paid. If the answer is "FTE hours" or "monthly retainer with FTE pricing logic," the partner has not internalized the buyer's preferred model. If the answer is "outcomes" or "named decision deliverables tied to a price," the partner has read the same trade press the buyer has.
The unoccupied third leg of the new buyer model is named accountable human counsel on the buyer side. Not a platform. Not an agent. Not a research tool with a dashboard. A named strategist who works with the buyer's agents, who interprets the output of the buyer's modular stack, and who signs their name to the recommendation.
The buyer has now publicly defined what they want. The holdcos are publicly building the wrong thing. The open indie coalition (Dept's Deptify orchestration layer, Horizon's accountable-major-independent frame) is publicly aligned against the holdcos but is structurally positioned at the orchestration layer, not the strategic-decision layer. The CI tools have just demonstrated their own category's structural exposure.
The strategy interpretation layer — the layer where a named human reads the agentic output, makes the strategic call, signs the memo, and answers to the buyer — is the only piece of the new architecture that is not currently being built by anyone. The buyer has named it. The holdcos cannot occupy it because their business model is incompatible. The open indies are not occupying it because their product is at the orchestration layer. The CI tools cannot occupy it because the buyer has now seen what happens when strategy data lives in someone else's database.
That is the wedge. The wedge has been there for three years. This week, the buyer named it in public.
Book a 30-minute demo. Bring a live question and watch the answer get built.