The meeting is tomorrow
Tomoko Hara has 30 minutes on the agenda. The CEO asked every division head to bring one big bet for the next fiscal year — something that changes how they operate, not just what they spend. Tomoko runs media and brand operations at Nova Motors. Her big bet is agentic advertising.
She's been thinking about this for months. She watched the agency present their "AI strategy" — it was mostly the same campaigns with a chatbot bolted on. She watched the programmatic team test another DSP feature that promised "AI-powered optimization." It made bids 4% more efficient. That's not a big bet. That's a rounding error.
Tomoko's pitch isn't about a tool or a platform. It's about changing how Nova's entire media operation works.
The problem she's solving
Nova Motors spends $180M a year on advertising. That budget flows through five channel teams — TV, social, programmatic display, search, and retail media — each with their own agency contacts, their own platforms, their own dashboards, and their own definition of success.
The TV team doesn't know what the social team is running. The retail media team at Walmart Connect has no idea what the programmatic team bid on that same shopper yesterday. Creative gets built weeks before media decisions are made, then force-fitted into whatever inventory the buyers found.
Every quarter, someone assembles a cross-channel report. It takes two weeks. By the time Tomoko reads it, the data is stale and the insights are obvious. She already knew social was underperforming. She didn't need a report — she needed the social team's agent to reallocate budget to CTV in real time, before she lost another week of the flight.
The real cost isn't waste. It's missed opportunity. Every channel team optimizes locally. Nobody optimizes globally. The signals from the retail media campaign (who actually bought the car) never flow back to the TV team (who drove awareness in the first place). $180M a year, and each dollar is lonely.
The pitch: three shifts
Tomoko doesn't pitch a protocol. She doesn't show JSON. She shows three slides.
From silos to a unified system
Today, Nova has five channel teams. Each one optimizes its own channel. They share a budget and a brand, but nothing else.
With agents, you have one system that operates across all channels simultaneously. The buyer agent that runs CTV knows what the social agent is doing. The programmatic agent knows what the retail media agent found. They don't optimize in isolation — they coordinate.
TV team runs reach campaigns. Social team runs engagement. Programmatic runs retargeting. No coordination. Same person sees the same ad five times on three platforms.
One system sees all channels. Agents sequence the journey: awareness on CTV, consideration on social, conversion on retail media. Frequency caps work across platforms, not within them.
Creative meets media in real time
Today, creative is built before media is planned. The agency produces 47 assets, and the media team tries to match them to placements after the fact. Half the assets don't fit. The other half are close enough.
With agents, creative adapts to placement. The brand identity — voice, guidelines, approved assets — lives in a machine-readable brand.json that every agent reads. When a buyer agent finds a sponsorship opportunity on a cooking show, a creative agent generates assets that match the context, the format, and the brand — all before a human touches it.
Creative team builds a 30-second spot. Six weeks later, media team discovers CTV inventory that needs a 15-second cut. Creative team is on another project. The spot gets cropped.
Brand guidelines are always available. When a new placement opens up, creative agents generate the right format, the right length, the right tone — and the brand rules are enforced at generation time, not in a review meeting.
Compound intelligence
Today, every campaign starts from scratch. The TV team learned something about SUV buyers in Q2. That insight lives in someone's head, or in a deck that nobody will find. It never reaches the programmatic team or the retail media team.
With agents, insights compound. Signals from one campaign feed the next. The retail media data that shows who actually bought a Nova EV flows back to the CTV agent that drove awareness. Over time, the system gets smarter — not because any one model improved, but because the agents share context across every touchpoint.
Each platform has its own model. Each model sees one slice of the customer. No platform knows what another platform learned.
Agents share signals across campaigns. Purchase data from retail media enriches CTV targeting. CTV reach data informs social sequencing. Each campaign makes every other campaign smarter.
What she's not pitching
Tomoko is careful about what she doesn't claim.
She's not replacing the agency. Pinnacle Agency is still running the campaigns. But instead of five separate briefs and five separate reports, Sam Adeyemi at Pinnacle sends one brief through a buyer agent that coordinates across all channels. The agency relationship is additive — they bring strategy, relationships, and creative judgment. The agents handle execution at scale.
She's not replacing measurement. Nova's MMM, incrementality testing, and attribution stack don't change. Agentic is a new way to execute, not a new way to measure. The protocol makes one thing easier: conversion events from retail media flow back through the system, so agents can optimize toward real business outcomes instead of proxy metrics.
She's not boiling the ocean. Tomoko's plan starts with one campaign across two channels — CTV and retail media. If it works, they expand. If it doesn't, they've learned something specific for the cost of one flight.
What happens in the room
The CEO asks the question every Head of Media dreads: "Why is this different from what the DSP told us last year?"
Tomoko's answer: the DSP optimizes within its own inventory. It doesn't know what's happening on social, on CTV, or in retail media. It's a really good specialist. What she's proposing is a system that coordinates the specialists.
"Think of it this way," she says. "We have five great musicians, each playing their own song. I'm not replacing any of them. I'm hiring a conductor."
The CFO asks about cost. Tomoko explains that the protocol is open — Apache 2.0, no licensing fees. The cost is integration and agency time, not platform lock-in. If it doesn't work, they turn it off. There's no three-year contract.
The CMO asks about brand safety. Tomoko pulls up Nova's brand.json on the screen — the machine-readable rules that every agent in the system reads before it makes a decision. Content restrictions. Geographic limits. Competitive exclusions. Not enforced by a human reviewing a spreadsheet after the campaign runs, but enforced at decision time, before a single impression serves.
She gets the green light.
The first campaign
Tomoko starts with the Nova EV launch. CTV for awareness. ShopGrid retail media for conversion. One brief, two channels, one buyer agent through Pinnacle.
Sam's buyer agent at Pinnacle sends the brief to both channels. On the CTV side, it discovers inventory on StreamHaus — Priya Nair's team curates the response, returning premium placements around auto-enthusiast content. On the retail media side, it discovers Daniel Park's ShopGrid storefront — in-store digital, sponsored search, and a product placement package tied to the EV's launch.
The buyer agent sees both responses side by side. For the first time, Tomoko's team can evaluate CTV reach and retail media conversion in one view — not in two separate dashboards reconciled in a spreadsheet two weeks later.
How the brief flows across channels
// One brief, sent to multiple sellers
{
"buying_mode": "brief",
"brief": "Nova EV launch. Target auto-intenders 25-54.
CTV for awareness, retail media for conversion.
Q1 2027, $2.4M total budget.",
"brand": { "domain": "novamotors.example.com" },
"filters": {
"channels": ["ctv", "in_store_digital", "sponsored_search"],
"start_date": "2027-01-15",
"end_date": "2027-03-31"
}
}When a shopper who saw the CTV ad walks into a ShopGrid store and test-drives the EV, that signal flows back. The CTV agent knows its impressions drove a real-world outcome. Next week, it shifts budget toward the programming that correlated with store visits. The retail media agent knows which in-store placements caught the attention of CTV-exposed shoppers, and bids higher on those next time.
By week three, the campaign is running differently than it started. Not because someone read a report and made a decision. Because the agents learned.
An always-on, omni-channel team of experts, in one system
That's what Tomoko pitched. Not a tool. Not a platform. A system that operates like the media team she wishes she could build — one where everyone talks to each other, creative adapts to context, and every campaign makes the next one smarter.
The protocol makes this possible because it gives agents a shared language. CTV sellers, retail media networks, data providers, creative platforms — they all speak AdCP. Tomoko's buyer agent doesn't need custom integrations with each one. It sends a brief and gets structured responses it can compare, coordinate, and learn from.
Six months later, the Nova EV campaign is running across four channels. The compound intelligence effect is real — CTV reach informs social sequencing, retail purchase data enriches programmatic targeting, and creative adapts to each context automatically. Tomoko's quarterly report takes a day to assemble, not two weeks, because the data is already unified.
At the next board meeting, the CEO doesn't ask "why is this different from the DSP." He asks: "What else can we run through this system?"
Tomoko smiles. She has a slide for that too.