Call my agent! Agentic AI begins buying media advertising

IBC

AI agents have the potential to automate media transactions at scale but today’s focus is less about fully autonomous workflows and more on experimentation and efficiencies.

article here

Agentic AI is rapidly gaining traction across connected TV (CTV) advertising, particularly behind the scenes where it is helping broadcasters and streaming platforms automate operational workflows, improve efficiency and accelerate new business models. However, according to executives from Bitmovin, Yospace and FreeWheel the industry is still in the early stages of understanding how agentic AI will fit into premium video buying and selling.

James Varndell, VP Product at Bitmovin, says operational adoption has accelerated significantly over the past six months.

“We're seeing agentic AI being used to speed up content preparation and enable more dynamic business models,” he explains.

A recent Bitmovin activation with Australia's Nine Entertainment is one example. The broadcaster wanted to introduce ad-supported streaming services and FAST channels alongside its subscription offering. Using AI tools, Nine analysed a library of approximately 17,000 assets to identify optimal ad break locations.

“Doing that manually would have taken a huge amount of time,” said Varndell. "By using AI agents, they were able to process the content much faster and launch their new ad-supported business significantly sooner.”

This shift reflects a broader trend across the industry.

“Rather than teams doing all the analysis themselves, agents are increasingly performing the analysis and then advertising teams review the output. It helps organisations move faster and operate more efficiently.”

According to Jeff Ellin, VP of Product Architecture at FreeWheel, the industry is still working through some fundamental questions around agentic buying and selling.

“People are trying to figure out what inventory they want to make available and what buyers actually want agents to do,” he says. “Do buyers want quick forecasts? Do they want to execute campaigns? Update campaigns? The idea of agents talking autonomously with each other is very real and will scale over time, but it's still very early.”

Testing the possibilities

FreeWheel, a Comcast-owned advertising technology company, participated in a proof-of-concept with NBCUniversal, agency RPA and Newton Research. Announced in January, the initiative demonstrated what the companies described as the first cross-platform agentic media buy spanning both linear television and streaming inventory. The initial execution centred on premium sports inventory, including NBC's Sunday Night Football.

Current implementations remain largely one-to-one. Buyers and sellers are building agents, connecting them and learning from initial deployments.

“Everyone is testing,” he says. “Some buyers are quite bullish on it, while others are more cautious.”

For live sports in particular, fully autonomous decision-making remains some distance away.

“The idea of agents autonomously agreeing a price for premium CTV inventory around something like SNF is not where the market is today,” Ellin says. “For lower-value, long-tail inventory, that might make more sense. For premium sports inventory, that's a very different conversation.

“Live sporting events are difficult to scale programmatically because of the speed and scale required," Ellin explains. “Sports actually became a really good use case because if I can sell inventory more directly and reduce reliance on SSPs and DSPs along the way, I may be able to solve some of those scaling challenges.”

Historically, a buyer might contact a sales representative to secure inventory. In the agentic workflow, a buyer interacted through a conversational interface, specifying the inventory requirements. The seller-side agent returned available inventory, and the buyer was able to complete the transaction through the automated workflow.

“The execution of the ad was exactly the same as it is today on Peacock,” says Ellin. “What changed was the way the transaction happened.”

New ad formats emerge

The real opportunity emerges when contextual intelligence is combined with live sports and emerging advertising formats such as squeeze-back ads. Unlike traditional overlays, squeeze-back ads resize programme content without obscuring it.

For example, AI could detect key moments during a football match, such as player substitutions, and trigger contextually relevant advertisements displayed alongside the live content.

ITV trialled the concept in during the 2026 Six Nations, where 20-second ads popped up during natural stoppages such as scrums. Even though this occurred just once per half reaction from fans was negative. The format was widely described as intrusive, with some supporters warning it risked damaging the viewing experience.

“Doing contextual analysis on live content introduces additional challenges, particularly around latency," Varndell says. “But we're seeing the technology mature quickly, and I expect we'll see more of these workflows move into live environments over the coming year.”

L-shaped squeeze-back ads positioned to the left or right of content and framed ads surrounding programme content are just two of a number of new standardised ad formats for CTV, being developed through IAB Tech Lab.

“The formats themselves have now been agreed," says Paul Davis, Head of Marketing at Yospace. “The work taking place this year is around creating the mechanisms required for trading, ad serving, measurement and reporting."

A key addition is dynamic QR code functionality, which introduces new attribution capabilities for television advertising.

“This allows advertisers to create a direct call to action and obtain a level of attribution that TV has previously lacked,” Davis explains.

The appeal of direct transactions

One of the biggest attractions of agentic buying is the possibility of creating more direct relationships between buyers and media owners.

“The way digital buying scaled outside of direct sales created a number of intermediaries," Ellin says. “The compelling idea for publishers is that if buyers can come directly to them at scale, some of those middlemen can potentially be removed.”

The result could be lower transaction costs, improved transparency and a greater share of media spend reaching publishers.

“Buyers traditionally want to get as close to the inventory source as possible,” he says. “You reduce the chances of fraud, technical issues and unnecessary hops in the supply chain.”

However, removing intermediaries also creates new challenges.

One example is frequency management. Today, buyers often use DSPs to control audience reach and frequency across multiple publishers. If transactions become entirely direct, each publisher would only manage frequency within its own environment.

“If I buy through a DSP across multiple publishers, I can control reach and frequency centrally," Ellin says. “If I go directly to individual publishers, I lose some of that control. These are exactly the kinds of problems the industry is trying to solve.”

Human oversight remains essential

Despite growing excitement around autonomous systems, most organisations remain risk-averse and want humans involved in decision-making.

“I don't see people simply letting agents run completely autonomously,” Ellin says. “Everything needs a human in the loop.”

AI can help scale monitoring processes to prevent unsuitable ad placements across vast content catalogues but even when AI agents are involved, “I still see humans making the final decision,” Varndell says. “The technology improves efficiency, but final sign-off typically remains with people.”

For many companies, practical workflow improvements are proving more attractive than full autonomous buying. Examples include campaign troubleshooting, delivery diagnostics and forecasting.

“Rather than having someone manually pull logs and investigate issues, AI can present what is wrong and suggest ways to fix it,” Ellin says. “Those are the kinds of use cases people are far more willing to adopt at scale right now.”

AI Connectors boost analytics

One area seeing particularly rapid progress is analytics.

Varndell highlighted recent developments around Model Context Protocol (MCP), which standardises how AI systems access and interact with data sources.

“AI systems can now connect directly to data through plug-ins and connectors," he explained. “We've built connectors for both ChatGPT and Claude.”

Bitmovin recently announced its availability as a Claude Connector, enabling users to query video and advertising data directly from within Anthropic's AI assistant.

According to Varndell, this dramatically lowers the barrier to data analysis.

“People who aren't data scientists can work with advertising data simply by asking questions rather than exporting spreadsheets and building complex formulas.”

The technology is particularly valuable for correlating data from multiple sources, such as video players and ad servers, helping organisations identify discrepancies and uncover performance insights more quickly.

Ellin says FreeWheel invested heavily in APIs allowing customers to build applications and workflows on top of FreeWheel's platform rather than relying entirely on its user interfaces and product roadmap.  

Recently, FreeWheel expanded this foundation through MCP-based infrastructure that enables AI agents to interact with platform services. The company is exposing these capabilities directly while also integrating AI-driven tools into existing user interfaces for customers not yet ready to build their own agents.  

“We want clients who are embracing AI to have complete flexibility in what they build,” Ellin says. “But we're also building similar capabilities directly into our platform for people who still prefer to work through a UI.

“Historically, software was built for departments,” he says. “What excites me is allowing individual users to create solutions that match exactly how they work.”

Two people in the same ad operations team may complete the same task differently. Through AI-enabled infrastructure, each could create personalised tools while still achieving the same business outcome.

Ellin says, “We're very bullish on what organisations and individuals will be able to create over time, beyond just agentic buying and selling.”