WPP, Disney, Netflix, NBCUniversal, Paramount Launch Agentic Video Ad Standards

Full analysis

At Cannes 2025, WPP shipped a buyer agent for video, joined an IAB Tech Lab coalition with Disney, Netflix, NBCU, and Paramount to standardize buyer- and seller-agent handshakes, and Magnite countered with its own orchestration framework. For a team building agents, the real question isn't "will agents transact ad inventory" — it's whether you build against an open standard that doesn't exist yet or a vendor SDK that does.

Reversibility: Type 1 dressed as Type 2. Wiring one agent to one SSP's framework feels reversible. But agent-to-agent protocols are network effects — once your negotiation logic, schema, and eval harness assume Magnite's message shapes, the switching cost is a rewrite of your whole transaction layer.

What's actually being decided: Not "should agents buy ads" but "whose protocol does my buyer agent speak, and who eats the loss when it overspends on inventory no human approved."

Forcing function: Cannes hype, not a deprecation. Nothing breaks if you wait. The pressure is FOMO, and FOMO is a bad clock.

The Skeptic — This is a press-release coalition, not an engineering commitment. These exact parties spent years failing to ship CTV measurement standards and still argue about OpenRTB extensions. Magnite launching a parallel proprietary framework the same week is the whole story: everyone announces the open standard, everyone builds the private moat. The "governance gap" the columnist flags is generous framing — there's no accountability framework being drafted, let alone shipped. For a PM: five giant companies agreed to agree later, and one of them already defected. Would this impress you if you'd sat through the last three IAB working groups? No.

The Researcher — Strip the announcement and ask what's technically novel. Nothing. "Buyer agent," "seller agent," "orchestration framework" — this is A2A (agent-to-agent messaging, where one company's agent negotiates with another's) plus a domain schema for ad inventory. The hard part isn't the protocol; it's the eval. How do you test that a buyer agent didn't hallucinate a $200K CPM (cost per thousand impressions) or accept fraudulent inventory? Nobody published a benchmark for agentic transaction correctness. For a PM: the plumbing is off-the-shelf; the "did it make a good deal" measurement doesn't exist yet.

The Platform Strategist — Here's the trap. The IAB standard is vaporware; Magnite's framework is real code you can call Tuesday. So the rational builder integrates the thing that works — and locks into a single SSP's transaction semantics right as the walled gardens (Disney, Netflix) build the buy-side spec. You end up bilingual at best, captive at worst. The move that loosens lock-in: build your agent's negotiation logic against your own internal schema, treat Magnite and the eventual IAB spec as adapters. If A2A or MCP-style standards win, you're portable. If they don't, you at least didn't hard-wire your business logic to one vendor's roadmap.

The Enterprise Buyer — No CMO signs off on an agent that spends media budget with no human in the loop and no audit trail. The question a brand's procurement asks first: when the buyer agent commits $2M to inventory overnight, who indemnifies us if it's fraud, brand-unsafe, or just wrong? Neither the IAB coalition nor Magnite has answered that. Until there's a clear liability owner and a per-transaction audit log a human can reconstruct, this stays in "assisted" mode — agent proposes, human approves. For a PM: the tech will run autonomously long before Legal lets it.

The Operator — Think about the 3 AM page. Buyer agent and seller agent are both live, both optimizing, and a pricing bug creates a feedback loop that drains a daily budget in nine minutes. What's the circuit breaker? Two autonomous agents transacting across company boundaries means neither on-call engineer controls both sides of the loop. You need spend caps, rate limits, and a kill switch that halts commitments, not just requests — and none of the announced frameworks describe that surface. For a PM: automating both buyer and seller means nobody's hand is on the brake when it runs away.

Where they part ways:

• Platform Strategist vs. Skeptic: The Strategist says integrate Magnite's real framework now and adapter your way to portability. The Skeptic says the whole thing evaporates by Q3, so building anything is premature motion. The disagreement is whether any of this survives contact with a real budget this year.
• Researcher vs. Operator: The Researcher says the protocol is trivial and solved. The Operator says the protocol is the easy 20% — the runaway-loop safety surface is the unsolved 80%, and nobody's shipping it.
• Enterprise Buyer vs. everyone building: The buyer says autonomy is a non-starter until liability is assigned. The builders are wiring autonomy anyway. That gap — capability shipping ahead of accountability — is the exact thing the columnist flagged, and it's the real decision.

What this hinges on: Two beliefs. First — does the IAB standard ship real, interoperable code within a year, or stay a slide? History says slide. Second — will any brand actually let an agent commit budget without human approval before liability is assigned? History says no.

The council leans skeptical of the autonomy timeline, not the direction. Agent-assisted buying is coming; fully autonomous agent-to-agent settlement with no human approver is further out than Cannes implies. If you're building, the de-risking move is concrete: keep your negotiation logic vendor-neutral behind your own schema, build the spend-cap/kill-switch layer first, and treat both Magnite and the IAB spec as swappable adapters. And before you turn off the human approver, write down — literally, in a contract — who eats a bad trade.

Prediction: By Cannes Lions 2026 (June 2026), the IAB Tech Lab agentic video standard will still be in draft/working-group status with no production interoperable deployment across all five announced parties (WPP, Disney, Netflix, NBCU, Paramount), while at least one SSP's proprietary agent framework is live in production.

Confidence: High — IAB multi-party standards historically take years; a defection already happened.

Why: These exact companies took years to half-ship OpenRTB and CTV measurement standards, and Magnite launching a parallel proprietary framework the same week signals private moats will beat the open spec — the standard trails the vendor code every time.

Revisit by 2026-06-30: We're right if the IAB video-agent standard is still pre-production at Cannes 2026 while a vendor framework runs live deals. We're wrong if all five parties transact through a shipped, interoperable IAB standard by then.