If this week in ad tech had a theme, it’s this: everyone finally agrees the system is broken, and their solution is to either build a new system, a new standard, or just quietly walk away from the old one while lighting a match behind them. The IAB Tech Lab is forming yet another council to fix programmatic, AI agents are multiplying like gremlins with competing “open” frameworks, and OpenAI is stepping into the ring with Google like it just realized ads are where the real money lives. Meanwhile, a publisher looked at the whole thing and said, in effect, “nah,” and deleted programmatic entirely.

What makes it all deliciously absurd is that every story is about control dressed up as progress. Control of standards, control of measurement, control of bidding logic, control of intent, control of the pipes, control of the money. The industry keeps saying “transparency” like it’s a group affirmation, but the behavior screams “who owns the stack wins.” And now AI is about to automate all of it, faster, cheaper, and at scale, which is either the fix or the final boss level of the problem. Either way, grab popcorn.

The IAB Tech Lab just launched a Programmatic Governance Council, which would be bold if the Interactive Advertising Bureau didn’t already have a programmatic council, plus a graveyard of working groups that have been “addressing transparency” since banners had drop shadows.

Here’s the real question everyone is politely dodging: is this innovation or an admission of failure? Because the same companies that turned programmatic into a $200 billion labyrinth of fees, duplicated bids, and “trust us” reporting are now… aligning again. The difference now is AI is about to scale the chaos. When the machines inherit your mess, suddenly governance becomes urgent. Funny how that works.

The Trade Desk drops Koa Agents. Microsoft Advertising fires back with its own AI stack. Meanwhile, frameworks are multiplying like bad startups: Open Agentic Kit, Agentic RTB, Ad Context Protocol, AgenticOS. Pick your fighter, then realize none of them want to lose.

Everyone says agents need to work together. No one is building like they actually believe that. This is not a kumbaya moment. It is a land grab with better branding. The punchline is brutal: the biggest thing blocking agent adoption will be the same fragmentation agents were supposed to fix. You automated the dysfunction. Congrats.

OpenAI moving into CPC pricing is not a tweak. It is a direct swing at Google’s core business. CPM was a nice sandbox. CPC is where performance budgets live and where marketers stop being polite and start asking hard questions.

Now OpenAI has to prove something uncomfortable: a ChatGPT click actually means something. Search works because intent is obvious. ChatGPT is still a mix of curiosity, research, and “let me double-check this robot.” Turning that into scalable, high-value traffic is the whole game. The new “thinking” image engine and agent workflows are the setup. The question is whether the demand shows up.

Index Exchange is letting DSPs run bidding logic inside its own pipes with Index Cloud. The pitch is efficiency. The reality is yet another line in ad tech getting erased in real time.

They swear they are not disintermediating anyone. Sure. And platforms swear they love open web competition. This is about control, plain and simple. Every player wants a bigger slice of a shrinking pie, and “collaboration” is just the polite word we use before someone gets cut out of the transaction.

CTV is messy, manual, and full of high-stakes deals that do not tolerate sloppy execution. That is exactly why it is becoming the first real test of agentic advertising. If AI can actually manage pricing, pacing, guarantees, and all the moving parts without breaking things, then maybe this works.

If not, CTV will expose it immediately. Nobody cares if your agent optimizes cheap remnant inventory. Can it handle premium dollars without screwing it up? That is the test. Everything else is demo theater.

The American Prospect walked away from programmatic ads entirely and explained why in painfully clear terms: surveillance, fraud, slow sites, privacy risks, and a revenue model that sends most of the money anywhere but the publisher.

This is what makes the industry uncomfortable. No optimization. No framework. Just “this system is bad, we’re out.” When a publisher decides the best move is to burn the stack down and ask readers for support instead, it raises a brutal question: who is programmatic actually working for?

A new browser-based attribution push is being framed as a privacy win. What it actually does is hand more control to the same platforms that sell the ads. Measurement, meet conflict of interest.

The risk is obvious. If attribution systems downplay quality environments and reward scale, money flows to cheap content and algorithmic sludge. The industry already knows about the halo effect. The question is whether a platform-controlled system will ever admit it. Spoiler: only if it accidentally helps them.

StackAdapt pitching ChatGPT ads at bargain CPMs is not disruption. It is market seeding. When a channel is truly must-buy, it does not need discounts and hand-holding to get budgets in the door.

This is the trial phase. The real test comes later when the novelty fades and buyers ask the only question that matters: does it outperform anything else I can buy with this money? If the answer is no, those CPMs are not a deal. They are a warning.

Hyundai talking about data-driven growth is fine, correct, and completely uninteresting. Of course it is data-driven. Everyone says that. The gap is not in strategy. It is in execution inside a system that is still opaque, fragmented, and increasingly run by AI layers marketers do not fully understand.

The real story is not that brands want better data. It is that they are making bigger bets in a system they trust less than they admit on stage.

Now. Having established that I am not a Luddite, let me tell you what is happening in advertising right now, because it is genuinely one of the funniest things I have covered in thirty years.

Two different groups of ad tech insiders are fighting over who gets to govern AI agents in advertising. Nobody asked either of them to.

In one corner, the IAB Tech Lab, longtime stewards of such beloved transparency initiatives as OpenRTB, ads.txt, and sellers.json, none of which fixed anything but all of which generated excellent PDFs. They have a shiny new Agentic Roadmap and a Programmatic Governance Council and the full institutional weight of being the people who have been failing to fix this industry since 2009.

In the other corner, AdCP, the scrappy upstart protocol positioning itself as the open-source alternative, pitched by people who looked at IAB Tech Lab and said "we can fail at this too, but faster and with better branding."

The trade press is calling this a standards war. I am calling it what it is: two committees fighting over who gets to charge you for certification on a technology neither of them invented.

Here is the part that kills me. Nowhere in this fight, nowhere, is the advertiser. Nowhere is the publisher. Nowhere is the consumer whose data is fueling the whole carnival. The "stakeholders" in both camps are the same cast of characters: platforms, DSPs, SSPs, holding companies, and the vendors who sell tools to all of the above. The house is arguing with the house about who gets to write the rules of the house. And we are supposed to care who wins.

I do not care who wins. You should not care who wins. The advertiser's best outcome in this fight is that both of them lose and somebody, anybody, builds an AI tool that actually works for the person spending the money, without asking permission from a council.

Because here is the actual promise of AI in advertising, the thing I am genuinely excited about. A good model, pointed at your own spend data, can tell you where your money went. For the first time in twenty years. That is a real capability. That is worth getting excited about. That is worth building.

And notice, notice, that neither standard is building that. AdCP is not building it. IAB Tech Lab is not building it. Because that tool does not need their certification. That tool does not need their protocol. That tool does not generate fees for their members. That tool is the thing their members are afraid of.

So they are racing to define "agentic advertising" on their own terms, before somebody outside the tent defines it on ours. Whichever one wins gets to issue the badges, run the conferences, and collect the compliance fees for the next decade. Whichever one loses will still have a LinkedIn page and a quarterly newsletter. Nobody is going out of business. Nobody ever does in this industry. That is part of the problem.

My advice, for the three advertisers and two publishers actually reading this instead of the eight hundred vendors: ignore both of them. Build your own tools. Hire somebody who understands LLMs. Point the model at your contracts, your spend data, your supply paths. Ask it the questions the council was never going to answer. You will learn more in an afternoon than a decade of "governance" will tell you.

The AI is real. The councils are cosplay. Pick the real thing.