The new agentic world is exciting. Not a day goes by that I don’t discuss something cool I’ve seen using Claude or realize that I can now vibe code an app to handle a task for my wife.
But with all new things, there’s a risk of neglecting our product management discipline in the bliss of our excitement. We saw it with crypto and when “blockchain” became a hammer in search of a nail at every turn. Without people pressing back on normie concerns – “How does this fit into my day to day? Would a normal database work as well?” – it’s easy for tech to get ahead of its skis.
Agentic tech is discussed in similarly grandiose terms, often at the expense of its more mundane actual value. One of the most talked about agentic ad use cases is direct sales.
But Google debuted something powerful at Cannes that significantly cuts into agentic AI’s direct sales value: Buyer Direct.
Still in beta with GAM publishers, Buyer Direct provides a direct booking option for agency accounts, which thus can transact with some of the benefits programmatic offers, but without DSP and SSP fees. And if Google has its way, Buyer Direct could put an end to all sorts of agentic direct sales opportunities while they’re still in the cradle.
Breaking down Buyer Direct
Agentic AI is quickly becoming enmeshed in all aspects of online advertising, from strategy to creative and media buying. But one of its major promises is to automate and improve the direct sales process, a task that has historically gone to order management systems. (These companies will rebrand as agentic platforms before you know it.)
Direct-sold deals, which are a major revenue line for publishers, are ripe for agentic innovation. Agents are great at deal sourcing because they can review publisher inventory so much faster than any human eyeballs; agents are also ostensibly superior negotiators as they lack emotion and personal preference. Besides, direct deals include vast amounts of manual work that can be automated and delegated to AI.
The problem is that agents are efficient players on an inherently inefficient battlefield. Publisher inventory availability is fragmented and piecemeal, even if the agents themselves are quite efficient.
Making the agentic framework work still requires an abstraction system that hooks into GAM or elsewhere via API/MCP (not like anyone has ever had any issues mucking about with GAM’s API). This system must also operate people’s emails or chats to document any back-and-forth and then handle reconciliation at the end of the campaign.
Still, there are inconsistencies in how publishers and people communicate about inventory, which leads to false starts and mistranslations between platforms. Even agents cannot fix the problem of unreliable integrations and API calls.
But Google can. Buyer Direct solves this agentic confusion because working with Google requires no communication or mediation between multiple platforms. There’s a single place where buyers can pick and choose inventory via Google’s agency interface, which they probably already use.
Competitive concerns
For direct-sold campaigns, I can’t imagine a world where Buyer Direct doesn’t give any other “agentic” product a run for its money. I frankly am not sure any outside agentic system can outperform Buyer Direct, considering almost the entire publisher market already uses GAM.
The important caveat is that Buyer Direct is primarily for display and online video ads. Non-Google agentic products might thrive in other channels because there isn’t the same level of ad server consolidation. However, display and OLV also happen to be the markets where Google just lost a massive antitrust battle. Google was declared an illegal monopoly in no small part because of its business-tying practices, where publishers were forced to use GAM to access Google’s proprietary ad network demand.
Now, Google may be realizing that it can invert the relationship. With 95% of the online publisher world already logged into GAM, the sell-side product can easily become its own source of liquidity. GAM doesn’t need AdX; it only needs an agentic-like interface, with Buyer Direct under the hood.
In other words, Google has seemingly figured out its plan B after AdX.
So yes, we should be excited about agentic AI. It promises to automate and improve all sorts of tasks. But even with the boundless potential at our fingertips, there’s no automating our way past the results driven by incumbency and superior architecture. We will need to innovate past what can be done with existing tools if we are going to create a thriving agentic ecosystem.
“Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.
Follow Gareth Glaser and AdExchanger on LinkedIn.
For more articles featuring Gareth Glaser, click here.
