There's this scene in a Spielberg film where Tom Cruise walks through a shopping mall and holographic billboards scan his retinas, calling out to him by name—"John Anderton! You could use a Guinness right about now." The scene now looks both impressive and unsettling, this vision of a future where advertising knows exactly who you are and what you want before you do. The film is Minority Report, released in 2002, and it imagined advertising infrastructure that was extremely intelligent and deeply integrated into daily life.
We're obviously not there yet with the retina-scanning billboards. But something potentially just as significant is happening in the less visible infrastructure that powers digital advertising today.
What AdCP Actually Is (And Who's Behind It)
In October 2025, a coalition of more than 20 advertising companies launched the Ad Context Protocol (AdCP), an open standard designed to standardize how AI agents communicate with one another in advertising. The founding members include Yahoo, PubMatic, Optable, Scope3, Swivel, and Triton Digital, with additional supporters like Magnite, Kargo, and The Weather Company joining at launch.
AdCP is built on top of Anthropic's Model Context Protocol. If you're into the technical weeds, you can check out the full specification on GitHub. Right now, they've got protocols for audience activation and media buying sorted out. Curation protocols are scheduled for Q2 2025, but we'll see if that timeline holds..webp)
AdCP is a protocol—a set of rules that allow different software systems to talk to each other. Built on Anthropic's Model Context Protocol, it provides a common interface for AI agents to exchange context and manage ad buying across platforms.
The whole point here is cutting through the insane complexity that exists today. DSPs, SSPs, agencies, tools — they can all integrate without spending months and tons of money on dev work. New systems, especially AI agents, can actually work without needing special APIs built just for them. And advertising data finally becomes more structured and standardized, which would be a nice change.
The Technical Foundation
Today's advertising industry operates through dozens of different APIs, formats and non-standard integrations. This fragmentation creates operational chaos. Each partner has its own requirements, its own endpoints, its own ways of transferring data. Implementing new tools takes weeks or months.
AdCP addresses this by enabling advertising platforms to speak the same language—a single standard instead of dozens of different APIs. The protocol creates a universal view of data about campaigns, budgets, bids, and ad formats, while enabling quick integration of AI agents that can automatically manage campaignsand provide transparent access to advertising context.
What makes this different from previous attempts at standardization? The people involved, for one thing. The founding members are from different parts of the programmatic advertising: supply-side platforms, sustainability measurement, audience analytics, audio advertising, and data collaboration tools.
In other words, this is an attempt to create shared infrastructure before everyone builds incompatible solutions.
The Shift Toward Agentic Media Buying
If you’re a media buyer, your day is fragmented. You log into multiple DSPs. You manually configure line items. You build audience segments through endless checkboxes. You set frequency caps that don’t always work across platforms. You struggle to combine reporting data from many different sources.
Research from mediasense and the World Federation of Advertisers found that 86% of companies rated automation as important for modernization, but satisfaction with current agency capabilities showed a 66-point gap.
This is where "agentic advertising" comes in. Instead of humans manually building campaigns, AI agents could handle the work through natural language instructions. You could just tell an agent, "Find me people interested in pet-related products." The agent would look through available data sources, negotiate with sellers' agents, and assemble the audience.
The main word here is "could." We're not there yet. But the infrastructure AdCP proposes would make it possible.
According to PubMatic's Kyle Dozeman, campaign planning, audience analytics, media activation, and troubleshooting are all components where agents are currently being tested, with the expectation that agents will eventually string together all parts of a campaign.
Understanding Buyer's Agents and Seller's Agents
The agentic model introduces two key participants that fundamentally change how advertising transactions work:
- Buyer's agents work for advertisers. They handle all the routine work that people usually do manually - picking inventory, adjusting bids, tracking how campaigns are performing, negotiating deals, buying placements. Basically all the stuff that takes up tons of time.
- Seller's agents represent publishers and media companies. They're controlling inventory availability, pricing, sales rules, creating private deals, managing buyer access through AdCP. All from the supply side.
Both types of agents talk to each other through special language: adagents.json files. It's a standardized format that lays out objects, what operations are allowed, how data gets exchanged, and how access gets verified.
.webp)
How AdCP Could Change Everything
Agent-to-agent transactions have become the standard way of doing things. Instead of publishers running real-time auctions where buyers bid against each other, advertiser agents would just communicate directly with publisher agents about what they need. The buyer's agent knows the campaign goals, the seller's agent knows what inventory is available. They negotiate directly through the protocol.
Here's how a typical transaction might work:
- A marketer makes a request to the Buyer's Agent: "I want to buy weather-relevant inventory from The Weather Company. Budget - $20k. Or 200m impressions"
- The Buyer's Agent converts this into technical requirements, constraints, purchase format, ideal CPMs, and audiences.
- The Buyer's Agent contacts the Seller's Agent via AdCP, asking questions like: What inventory is available? What are the minimum CPMs? Are there cross-platform packages? What audience segments are available? Is it possible to create a private deal?
- The Seller's Agent responds with inventory structure, format descriptions, available deal options, minimum CPMs, restrictions, and access rules. It’s all in a single AdCP format.
- The AI agents negotiate deal terms automatically, establishing CPM, budget cap, audience parameters, formats (Native/Display/Video), delivery windows, frequency cap, and volume guarantees.
Premium supply gets better representation. This has been a chronic problem in programmatic — high-quality publishers really struggle to differentiate themselves in auction environments that just reduce everything down to CPMs. When agents can actually query for specific content contexts and audience qualities through standardized protocols, premium inventory becomes way more discoverable.
Barriers drop for smaller buyers. Right now if you want to do effective programmatic buying, you need technical expertise and usually you need agency partnerships. But if natural language interfaces backed by these agentic protocols actually become viable, smaller advertisers could potentially access the same infrastructure that major brands use. That would be a pretty big deal.
Inventory outside traditional marketplaces becomes accessible. The agentic ad channel doesn't have to be limited to just programmatic stuff. Campaigns could include custom components that aren't available programmatically, or even offline inventory. Take, for example, podcast sponsorships, newsletter placements, emerging formats in gaming and connected TV — all of it discoverable and transactable through the same protocol.
The Realistic View: Evolution, Not Revolution
Every new standard faces absolutely brutal adoption challenges. Look at OpenRTB — that's the protocol powering most programmatic advertising today. It took YEARS to achieve widespread adoption. And even now, it's still implemented inconsistently across the industry.
AdCP is going to face similar hurdles, maybe worse. Will enough publishers actually adopt it to make it valuable for buyers? Will buyers trust AI agents to make decisions that currently require human judgment and experience? Anne Coghlan from Scope3 pointed out that "right now, there is always a human in the loop." And honestly, that's probably staying true for quite a while.
The most likely outcome is some kind of hybrid future. Traditional programmatic isn't just going to disappear overnight. Instead, AdCP-based agentic tools will probably coexist alongside conventional DSPs and SSPs. Each handling different types of transactions. Early adopters might get some efficiency advantages, but real transformation - if it even happens — is going to unfold gradually over years, not months.
There's another thing worth mentioning. PubMatic's infrastructure experts noted that agentic AI systems need 20 to 30 times more compute power than generative AI systems. And generative AI already demands way more than current programmatic systems. So this is a fundamental infrastructure shift that most legacy platforms really aren't equipped to handle.
In Minority Report, all the technological infrastructure just... existed. The mall had it, the transit system had it, everyone was connected and it all worked seamlessly. But in reality? Building that infrastructure is the actual hard part. AdCP is essentially trying to lay the foundation before everyone goes off and builds incompatible versions of the same thing, which is what usually happens in this industry.
Where SmartyAds Fits Into This Picture
As an independent DSP, we've always taken what I'd call a practical approach to industry standards. We adopt innovations when they actually deliver proven value to media buyers.
With AdCP specifically, our approach right now is informed observation. The protocol is technically sound. The coalition behind it is credible — these aren't random startups, they're established players. And the problems it's trying to address? Fragmented workflows, rising operational costs, inefficient discovery of inventory and audiences — these are all very real, very painful problems.
Our commitment is to interoperability. As standards like AdCP mature and actually prove themselves through real implementation and measurable improvements in campaign performance, we can and will integrate them in ways that’ll enhance audience discovery, operational efficiency, and transparency for our buyers.
That might mean supporting AdCP-based audience activation at some point. Or integrating with agentic buying tools if they prove valuable. Or connecting to new inventory sources that the protocol makes available.
Our goal is to stay flexible and adaptable. Because let's be honest — most innovations in adtech don't pan out. Media buyers need partners who can tell the difference between something that works in practice.
What Comes Next
AdCP represents what I'd call a considered bet on where advertising infrastructure is headed. Whether it becomes foundational or just remains a niche thing depends entirely on adoption rates, implementation quality.
What IS certain though is that the underlying problems aren't going anywhere - they're actually getting worse. That Mediasense and WFA research showed that 92% of companies believe speed and agility are important, but only 31% are satisfied with how their agencies are delivering on that. That's a 61-point gap, which is pretty significant. Programmatic spending just keeps growing year over year, but buyer satisfaction with the actual process hasn't kept pace at all.
So the question really isn't whether we need better solutions. We obviously do. The question is whether AdCP specifically provides those solutions, and whether the industry can actually coordinate effectively enough to make standardized agentic advertising work.
For now, it's definitely worth watching closely. The coalition is substantial, the technical approach seems sound, and the problems they're addressing are legitimate. Adam Broitman from McKinsey said that "AdCP gives the advertising industry a chance to establish shared technical infrastructure and standards that aim to accelerate progress."
The media buying landscape has weathered plenty of supposed revolutions that ended up being just incremental evolutions. Maybe AdCP is different. Or maybe it becomes another layer in the already complex infrastructure we all have to navigate daily, delivering some benefits but also introducing new challenges. Too early to tell really.
Either way, the conversation about making media buying less manually intensive and more intelligently automated isn't going away anytime soon. AdCP is one possible answer to that challenge. Time will tell if it's the right one, or just another attempt that falls short.
In Minority Report, the future of advertising just worked perfectly. The mall had it, the transit system had it, everything was connected and functional and seamless. But that's because Spielberg's film conveniently skipped over all the boring parts - the years of building protocols, testing agents, arguing over standards, watching half the implementations fail because nobody could agree on anything. We're living through those boring parts right now. AdCP is one piece of that larger puzzle.
Maybe it becomes the foundation everyone eventually builds on. Maybe it ends up as just a footnote in some future article about "standards that almost happened but didn't quite make it." Honestly, it's way too early to say which way this goes.
