Connected TV won’t scale unless it ditches old programmatic tech
Opinion – The Future of TV: Global Series
AI agents will restore control to advertisers and publishers through programmatic CTV by boosting content transparency. Converge’s CEO Ian Maxwell believes it will open the door to more long-tail advertisers.
When CTV publishers began laying the pipes for digital advertising, they looked to the open web for inspiration.
This seemed sensible; after all, there are decades of innovations that could be pulled from the ad tech industry. But by the time the first programmatic transaction put an ad on the telly, the infrastructure that got it there was already outdated.
By replicating the open web model, with DSPs on one side, SSPs on the other, and a raft of questionable intermediaries in between, the burgeoning CTV market also replicated its failures: black box transaction logic, a ballooning tech tax, and a supply chain that creates too many places where fraudsters can hide.
AI can now sweep out the clutter.
I don’t mean a chatbot overlaid on the same old buying platform, or a generative AI that summarises your campaign reports. I mean autonomous agents that buy, sell, target, and verify on behalf of their respective parties.
Instead of trading through a Rube Goldberg machine of creaky programmatic tech, publishers and advertisers can transact directly with lightweight, transparent AI agents.
CTV deserves better than open web leftovers
Let’s start at the top of why CTV should put legacy programmatic tech on the scrap heap: the fraud.
If you’ve peeked under the bonnet of the CTV supply chain, you might have noticed some pretty loose definitions of what counts as TV. There’s a rogue’s gallery of resellers out there packaging any video that could potentially be played on a TV as CTV inventory.
I’ve personally never loaded up a web browser on my TV to watch an embedded video, and I doubt many others do either, but apparently that’s good enough for unscrupulous resellers to pass off on unsuspecting buyers as bona fide CTV inventory.
Perhaps there’s enough plausible deniability for such mis-selling to stay on the good side of the law, but common sense says if someone wants to buy a CTV ad, they mean the big screen in the living room that gives the full premium advertising experience to a captive audience, not a video that pops up halfway through an article while someone’s browsing the web on their train to work.
If you’re an agency with the requisite spending power, or a publisher whose ad slots are in high demand, you can dodge fraud by buying and selling directly, or you might trade exclusively through strictly curated supply chains.
That’s fine for the select few, but bad for the health of the CTV ecosystem as a whole.
Why? Because there’s lots of money in the long tail.
Search and social media rake in the cash not because they’re premium brand-building environments, but because smaller advertisers can easily target relevant audiences en masse.
And because such buyers purchase smaller volumes, they also pay a higher average CPMs than big brands that buy in bulk.
CTV won’t reach its full profit potential unless it’s broadly accessible, and this won’t happen unless buyers can trust their spend will go to the right place.
I don’t like giving Big Tech credit, but this is something the folk at YouTube understand: within one ad pod, you might see a glossy, celebrity-packed Nike campaign alongside a promotion from your local Italian restaurant.
Agentic trading is all supply, no chain
CTV buyers and sellers can trust AI agents because they don’t work for anyone else.
Instead of hoping that loose signals used to define content, device type, and audience are carried accurately through a supply chain they can’t scrutinise, they can use agents trained to filter the noise on their behalf.
The process is very simple. An AI agent is set up to monitor millions of programmatic transactions to recognise signals (such as session length, device type, show titles, and environmental cues) that indicate the person on the other end of the screen is having a true CTV experience.
Once confirmed, to make sense of the inventory, these signals are run through an enrichment pipeline that pulls structured descriptions and contextual metadata from entertainment databases, live web sources, and more.
The result is a clean, consistent, natural-language description of what the content is actually about.
Once the AI agent’s learning is complete, it can be deployed at scale without requiring any questionable third-party vendor input.
This is where raw inventory is understood for what it is, and downstream targeting systems can then match against this data rather than flying blind.
This same learning process can be applied to audience targeting, ad verification, brand safety, and many other functions vital to accessible programmatic trading, which are currently the domain of margin-extracting intermediaries.
To understand how agents could help you, think: How would you tackle a task if you had near-infinite time and energy to dedicate to it?
How AI agents operate
That’s effectively how AI agents operate. You give them instructions, set them off to learn how to do the task in a test environment, then they can be relied on to execute at scale with minor tweaks along the way to keep them on the right track.
In place of a long road between buyer and seller dotted with middlemen demanding tolls from anyone who passes, both parties deploy teams of purpose-built agents trained for specific functions.
An agentic market is an open market.
That towering tech stack that always feels like it’s about to topple over? No longer necessary. Those fraudsters duping buyers with mis-sold inventory? Nowhere to hide.
The percentages shaved off every transaction by programmatic gatekeepers? Keep it for yourself.
AI agents give control back to the advertisers and publishers who keep money flowing through the CTV market, which is exactly what the channel needs to achieve its full growth potential.
Ian Maxwell is the CEO at Converge.