• by Maarten Albarda , Featured Contributor, May 29, 2026

Does this sound familiar to you? A revolutionary new technology arrives, promising to democratize access, simplify the complex, and deliver performance at the push of a button. We all pile in.

Then, a few years later, the Association of National Advertisers releases a study showing that for every dollar you spent, only about 40 cents actually reached a consumer’s eyeballs. The rest? It vanished into the ad-tech tax, a maze of demand-side platforms, supply-side platforms, and data "enrichment" fees.

The industry, me included, has written extensively about the black hole in programmatic advertising. And as I look at the recent moves from OpenAI, agencies and ad-tech providers, I’m getting a distinct sense of déjà vu -- and not the good kind.

OpenAI has officially opened its self-serve advertising interface. It is targeted at everyone from the giant blue-chip brands to your local car wash or dry cleaner. It’s the classic Meta and Google playbook: Build the platform, create the self-serve tools, and then let the money roll in.

But here’s where it gets a bit murky. Even as the interface goes live, agencies and ad-tech firms are already positioning themselves as the "essential edge." They are claiming that you shouldn't just buy ads from OpenAI directly. No, no. You need to go through outside firms to get "deeper data integrations" with “real-time reporting and optimization."

Sound familiar? It should. It’s exactly how the programmatic black hole was constructed.

We are currently at a defining inflection point. According to the Q3 2025 ANA Programmatic Transparency Benchmark, we’ve actually made progress in regaining digital ad dollars. Working media dollars on web and mobile jumped to 47.1%. That means we finally managed to recover about $13.6 billion in value that used to just... disappear.

So, after a decade of fighting to claw back that 11-point gain, why on earth would we allow a new version of the same problem?

We are entering the era of agentic AI. These systems don't just answer questions; they plan and execute multistep workflows autonomously. They can run a media buy or process invoices end-to-end. It’s powerful stuff. Add to that the emergence of agentic to agentic media trading that is being tested (and celebrated by the techies), and you can understand my concern. If we let agentic middle tech sit between our budgets and the platforms without absolute transparency, we are allowing a new, higher-tech black hole.

We’re already seeing embedded AI quietly appearing in software we already own, often with updated terms that grant vendors rights to our data for model training. It happens at renewal, often without a new contract trigger, and suddenly your data is feeding someone else's model.

So, what should you do?

When an agency or third-party tech platform says you need to buy OpenAI through its proprietary agentic layer for an edge, ask them to prove the ROI before you sign, not after. If they can't quantify the value of the AI feature separately from the base platform, the "capability" is likely just a markup in disguise.

And while you’re at it, update your master services agreement to explicitly prohibit vendors from using your proprietary data (pricing, spend, strategy) to train their models without written approval.

We have a chance to get this right this time. Let’s not spend the next 10 years trying to recover value we should have never given away in the first place.