The Programmatic Stack Has Become Too Complex. AI Could Finally Simplify It

By Nic Baird, CEO and Co-founder of Koah

For three decades, digital advertising has added technology layers in the name of efficiency. The result is a programmatic ecosystem that can be remarkably sophisticated — and remarkably inefficient.

A 2023 ANA study tracing $123 million in programmatic ad spend found that only 36 cents of every advertiser dollar reached the consumer. The rest was absorbed by fees, fraud, non-viewable inventory and made-for-advertising sites. The complexity of the programmatic supply chain has become a problem in itself.

The industry didn’t get here by adding unnecessary tools and platforms. Most were built to solve legitimate problems. Ad networks addressed unsold inventory. DSPs and SSPs created scale and automated auctions. Header bidding gave publishers more control. Verification and identity infrastructure emerged to address fraud, privacy and measurement concerns. But these solutions have tended to persist long after the problem itself changed. By 2026, a publisher can find itself managing a stack of roughly 17 layers depending on the surface, from ad servers and SSPs to identity providers, clean rooms, verification, CTV infrastructure and increasingly AI-driven decisioning.

The question for the next era of advertising shouldn’t be: What technology should we add? It should be: What technology can we remove?

Everyone is taking a cut

The open web remains one of the most middleman-heavy environments in advertising. A typical transaction can move from a publisher’s ad server through a header-bidding wrapper, multiple SSPs, an exchange, a DSP, verification and an advertiser’s ad server before an impression reaches the consumer. Each individual layer can take a relatively small percentage. Collectively, those percentages compound.

The result is an ecosystem in which publishers often have less visibility into the economics of their own inventory than they should. According to Jounce Media’s 2025 supply-chain analysis, the average RTB-enabled publisher was integrated with 26.8 SSPs and authorized 16.7 partners to resell its inventory. More partners do not necessarily mean more value. In many cases, they mean more duplication, more fees and more opportunities for the dollar to leak out of the system.

The winners control the surface, the data or the outcome

But this isn’t the only model. The companies succeeding today and capturing the most value look very different from the traditional adtech middlemen.

Consider AppLovin. Its advantage isn’t simply that it has a better bidder. It owns critical pieces of the mobile ecosystem, including its SDK, first-party behavioral data and advertising infrastructure. Its AXON system can then use those signals to optimize toward advertiser outcomes. Amazon, Meta and Google have built similarly powerful positions by combining access to users, first-party data, advertising surfaces and increasingly sophisticated AI-driven decisioning.

The lesson for publishers isn’t that every company needs to become Amazon or AppLovin. It is that control matters. The strongest advertising businesses increasingly combine a scarce asset, such as a valuable audience, authenticated user base or proprietary inventory, with data and decisioning that make that asset more valuable.

That is a very different model from simply inserting another intermediary into the supply chain.

AI could be the reset button

This is where AI presents an opportunity. AI isn’t simply another technology layer being added to programmatic advertising. It is changing the interface through which consumers discover information, evaluate products and ultimately make decisions. Search, shopping and recommendations are moving from lists of links toward conversational and increasingly agentic experiences. As that happens, entirely new advertising surfaces will emerge inside AI interfaces. That creates an opportunity to build the next generation of advertising infrastructure differently from the start.

Instead of recreating the existing stack inside AI, the industry can prioritize fewer intermediaries, first-party intent, outcome-based optimization and formats designed specifically for the behavior taking place. A sponsored product in a traditional search result is built around a search query. A sponsored recommendation inside an AI conversation could be built around a decision, a constraint or a specific task. The format should match the behavior.

What publishers should do now

For publishers, the path forward is surprisingly straightforward.

First, reduce unnecessary complexity. A smaller number of high-quality, transparent partners can be more valuable than dozens of overlapping integrations. Second, treat first-party data as an asset whose value is rising fast. As third-party identifiers disappear, logged-in users, subscription relationships, purchase behavior and declared intent are increasingly valuable. Third, get closer to advertisers. The fewer intermediaries between premium inventory and the brands buying it, the more value can remain in the ecosystem. Fourth, optimize toward outcomes. The next generation of advertising will increasingly be judged by what an impression accomplishes, not simply whether an ad  was served.

Finally, build new formats around new behaviors. The biggest opportunities in AI, CTV and commerce media will not come from adapting yesterday’s banner, video or sponsored-product formats. They will come from creating advertising experiences native to the way consumers interact with those environments.

The programmatic industry spent 20 years building infrastructure to solve the problems of scale, targeting and automation. That infrastructure created enormous value, but it also created enormous complexity. AI gives the industry a chance to rethink the architecture. We don’t need companies that add another technology to the stack. We need ones that can remove layers that are no longer useful.