Agentic Advertising is for Allocation

“OpenRTB is a protocol for day trading; AdCP is a protocol for investing” - Benjamin Masse

Programmatic was invented to solve a problem with remnant ad impressions. How could publishers know, for a particular impression, which ad network would pay them the most money? We still struggle with this question, which is why mediation companies like CloudX are raising huge rounds and AppLovin is so valuable.

Premium publishers and brand advertisers have a different set of questions, much more akin to other capacity-constrained markets like hotel rooms and airplane tickets: given a limited amount of inventory or budget, how do we allocate to maximize our returns?

Quants vs PMs

In the finance world, quants are the maths people. They build complicated algorithms and obsess about how to get an edge when they execute them. Portfolio managers are the investors who evaluate fund managers and companies to decide how to allocate their clients’ money.

RTB is for quants to put a complex algorithm on the edge of the internet. AdCP is for portfolio managers to allocate our money to maximize returns. This allocation problem requires a set of tools to:

• Discover investment opportunities

• Transact with funds and companies

• Monitor performance

Programmatic advertising is built for quantitative buying of individual impressions. This solves a real problem, just as on Wall Street there are entire companies built around high-frequency trading and quantitative finance.

On the flip side, the vast majority of finance - and advertising - operates on longer timeframes - from venture capital and private equity to mutual funds and ETFs. Portfolio managers can and do allocate to quants, and often make a ton of money doing so, just as agencies allocate to programmatic trading desks.

Programmatic, like high-frequency trading, solved scale through standardization: if every impression or ticker symbol looks the same, machines can trade them. But the majority of advertising, like finance, isn’t standardized—it’s negotiated. Sponsorships, upfronts, custom integrations, premium placements. These require judgment, context, and back-and-forth. Until now, that meant humans, which meant cost, which meant you could only work with a handful of partners.

Agents change the economics. They can interpret a rate card, negotiate terms, adapt to a publisher’s specific workflow, and execute—without requiring every seller to conform to a single protocol. This is why an advertiser can go from buying 3 platforms to 20 publishers without tripling their team. The constraint was never technology. It was the cost of complexity. Agents collapse that cost.

Consider: when ChatGPT adds advertising, will it share user data with the open market to solicit programmatic bids? Or will it offer advertisers an AdCP-enabled API where it uses its proprietary data to drive outcomes? The answer is obvious—and it’s the same reason Snap and Pinterest don’t open up to RTB. Programmatic would commoditize their differentiation: their ad formats, their first-party data, their ability to optimize toward outcomes rather than impressions. Same for retail media networks, where the real value is closed-loop measurement and optimization from POS data—something RTB can’t price.

RTB’s objective function is the impression. But advertisers don’t buy impressions; they buy outcomes. Allocation lets media companies compete on what they’re actually good at: turning advertiser dollars into consumer eyeballs and purchases.

Allocation vs Valuation

The fundamental question that a programmatic trader has to answer is “what is this impression worth?”

The fundamental question that a portfolio manager has to answer is “how much should I allocate?”

This confuses ad tech people who have grown up in programmatic. Why do agencies and marketing folks not obsess about the nitty gritty of targeting, audience data, and supply paths? Why don’t they care about the inefficiencies of programmatic?

What would you rather have: a fund that returns 20% a year and charges you 6%, or a fund returning 5% a year that charges you 1%?

What this means will be painful to ad tech people: at the end of the day, every supplier is just another fund (ad network!) to allocate to. In finance terms, everything looks like an investment. All that matters is what the return is (and if you get sophisticated, correlations and risk).

Portfolio managers don’t care what you paid for an impression. They do care whether, in aggregate, the impressions they buy drive business outcomes.

Portfolio managers don’t care whether your supply chain is efficient. They do care about the overall effectiveness of their media investment.

Portfolio managers have a simple question: does a given media investment drive people to purchase my product (short term), consider my product (medium term), or like my brand (long term)? And how does this investment compare to other places I could put my money?

Agentic is for Allocation

Again from Benjamin Masse:

Agentic advertising doesn’t just move us from trading to allocation, it also breaks the ID-only, short-horizon worldview baked into real-time bidding. RTB isn’t just day-trading tech: it’s a snapshot protocol that can only reason about users and moments.

That’s why both walled gardens and MFA sites won: they optimized exactly what the protocol could see, while walled gardens could additionally layer their own internal yield management on fully traceable users.

Agentic advertising closes that intelligence gap by letting the market reason about things PMs actually care about long term:
– narrative horizon (time, repetition, shelf-life)
– institutional trust (environment as part of the message)
– relationship provenance over surveillance
– outcome feedback over months, not clicks

In that sense, AdCP really provides the allocation / reasoning layer — which also makes it a clean on-ramp for GenAI/LLM-native ad systems entering an ad-supported world, and a more viable long-term paradigm for premium publishers on the open internet.

AdCP works for linear TV and CTV. AdCP works for out of home and digital out of home. AdCP works for social platforms and media companies. It works for premium inventory and bundles as well as remnant.

The mission of AdCP is to be a protocol for ALL of advertising, not just for the part you can auction in real time. As Terry Kawaja says, the Total Addressable Market here is massive - and the opportunity to accelerate the global economy.

Agentic advertising increases ROAS

Portfolio theory tells us that we need to look across a set of investments to make optimal decisions because 1) many investments are correlated with each other, 2) investments operate on different timeframes, and 3) we can’t predict the future.

These principles clearly apply to advertising. Ad buys overlap across channels and within channels. If you’re saturating St. Louis with TV buys, that has a major impact on your digital spend. Upfronts are an illiquid investment, as are many forms of sponsorship. And we can’t seem to figure out backward-looking measurement, much less a model for prediction.

Assembling a media plan is building a portfolio. Agencies have investment teams that evaluate suppliers and negotiate with them.

Agents are optimal for complex decisions that involve reasoning, interaction, and context, whether on the buy side or sell side. AdCP is built to help publishers, platforms, and agencies maximize portfolio value through negotiation and allocation.

Will AdCP improve my returns on advertising? Can it:

• Surface more and better ad products for advertisers to invest in

• Lower transaction friction and costs across a diverse portfolio of products

• Make more effective allocation decisions across the portfolio

Today, most advertisers buy effectively from 3-5 platforms because execution costs scale with complexity. Instead of requiring a team to scale, agents make it a workflow. Diversify to 20 suppliers and portfolio theory does the rest.

Agentic advertising provides automation, tools, and intelligence to parts of the industry that haven’t been served well by ad tech. Agentic advertising is the allocation layer for a $1T industry that could be $2T if we improve discovery, execution, and allocation at scale. This is the performance play: not adding more compute to tweak individual bids; reallocating spend at a macro level to find billions of dollars of opportunity.