Agentic AI 对广告发布方的真正考验不是定价，而是库存还能否获得出价

Monday, November 9th, 2026 – 12:35 am

The agentic skeptics have the better argument right now.

Anthony Katsur, who runs IAB Tech Lab, expects “several false starts” and years of experimentation and alignment across platforms, agencies and publishers. Razorfish called the mood around agentic AI inevitable – but incremental – at CES back in January.

And Digiday’s research in January found that 54% of 142 brand and agency professionals say their companies don’t use agentic AI in their workflows at all. That is the buy side reporting on itself.

I have spent 20 years watching this industry call a shift imminent about once a year. Mobile was called inevitable around 2007, but it did not pass desktop in US digital ad spend until 2015. Cookie deprecation was announced in January 2020 and abandoned in July 2024.

Anyone telling publishers they are already behind on agentic AI is selling something.

However, there’s a fundamental difference between agentic AI and the “inevitable” industry shifts that came before it. Each of those shifts required somebody to build and pay for new integration, and the parties who had to pay disagreed about what to build. The variable is sunk integration cost.

Agentic is the first shift where buyers already paid that bill.

Agentic AI doesn’t face the same barriers. Fully autonomous media buying may still be years away, but AI agents don’t need new pipes or industrywide standards to start evaluating inventory, informing bidding decisions and optimizing spend. They can operate through existing programmatic infrastructure, using integrations the buy side has already built and paid for.

Money is already moving

On April 8, 2026, Microsoft ended a relationship with Dentsu that ran for more than a decade and moved its global media planning and buying to Publicis. That mandate carried a price tag the trade press puts between $700 million and $1.2 billion. The media appointment sat inside a larger announcement about embedding agentic AI across the marketing workflow: agents on Microsoft Fabric, informed by Epsilon, deciding which segments matter and deploying media against them.

Asked why Publicis keeps winning without pitches, Arthur Sadoun said clients now understand “it’s about identity.” Publicis also agreed in May 2026 to acquire LiveRamp, a deal the FTC cleared in October subject to a 5-year neutrality order. Once it closes, one of the two largest buyers of publisher inventory will own more of the identity layer that its own agents reason over.

Those agents buy media. Media clears in auctions. Publishers get paid in those auctions.

New logic, same infrastructure

The industry’s experience with bid shading shows how quickly new buying logic can reshape auctions.

Exchanges began moving to first-price auctions in 2017. Google announced Ad Manager’s move to first-price in March 2019 and finished four months later. Bid shading, a strategy for avoiding paying too much in first-price auctions, spread across SSPs and then DSPs within months via existing integrations, without new infrastructure or industrywide standards.

Publishers, meanwhile, were left to understand bid shading’s impact through their clearing-price reports and adjust their pricing floors accordingly.

Call bid shading machine learning rather than agentic if you prefer. But what’s different this time with agentic is what the logic decides. Shading optimized the price of a bid you were already making. But identity-informed agents decide what inventory is worth a bid at all.

The distinction matters. Publishers risk losing opportunities before their inventory even reaches an auction.

The risk is missing the auction entirely

Think of agentic logic as being divided into two halves. IAB research underscores the divide between AI-powered decision-making and autonomous execution. Among 360 digital video buyers surveyed this February and March, 96% see a role for agentic AI in programmatic and RTB, and 63% expect that role to be an optimization layer that improves bidding.

But current adoption, in IAB’s own words, is centered on decision support rather than the purchase decision itself. AI-based deal execution and negotiation for video campaigns sits at 28%. Forty percent of buyers require human approval before an agent acts. That is the shape of it: Agents run freely where no money is committed and get gated where it is.

When Katsur predicts those agentic false starts, he is describing the execution half, and he is right about it.

But the decision-making half is already gaining traction. Among video buyers spending $50 million or more a year, 63% are using or planning to use agents for inventory discovery and evaluation, against 30% of smaller buyers. The largest budgets in the market are pointing agents at deciding which inventory is worth considering.

The exposure is not whether agents will pay a premium for your inventory. It is whether your inventory survives evaluation. Fail that, and you will not be outbid in some future auction. You will not be in the running at all.

Which means the work is the same work it has always been, and it pays now. Publishers don’t need to wait for agentic standards to mature to prepare. Much of the necessary groundwork can also improve performance in today’s auctions.

Here are three steps to take, none of which require knowing which standard wins:

Create stable identifiers across your properties
Structure first-party attributes so bid requests carry clear context
Authorize data access explicitly, and get the terms in writing. Bid-request signals can be visible even to bidders who don’t win, which makes authorization a contract question rather than a technical one

These steps can strengthen inventory evaluation and match rates today, regardless of how quickly autonomous buying develops.

Remember: In 2019, the publishers who understood their own floors before the auction changed did fine. The ones who only learned about the change in buy-side decisioning from their revenue reports did not.

With agentic AI, waiting for fully autonomous transactions could mean missing the more consequential shift already underway.

“The Sell Sider” is a column written by the sell side of the digital media community.

Follow GrowthCode and AdExchanger on LinkedIn.

For more articles featuring Jonathon Shaevitz, click here.