Podcast Summary: "Inside NBCUniversal’s Test to Use AI Agents to Sell Ads Against a Live NFL Game"

Podcast: The Digiday Podcast
Host(s): Kamiko McCoy (Senior Marketing Reporter), Tim Peterson (Executive Editor of Video & Audio at Digiday)
Guest: Ryan McConville (Chief Product Officer & EVP of Ad Products and Solutions, NBCUniversal)
Date: February 3, 2026

Main Theme & Purpose

This episode delves into NBCUniversal’s groundbreaking proof of concept using agentic AI to buy and sell advertising inventory—specifically for a live NFL playoff game, one of the most valuable ad opportunities in TV. The discussion addresses the current state of AI agents in ad transactions, the intricacies and challenges of automating the traditional TV and streaming ad sales process, and the implications for buyers, agencies, and viewers.

Key Discussion Points and Insights

1. The State of AI Agents in Ad Transactions

[00:34 – 02:21]

• Before NBCUniversal's experiment, the consensus in the industry was that AI agents were limited to planning, reporting, and data analysis, but not actual ad transactions.
• NBCUniversal, with partners RPA, Noon Research, and Freewheel, set out to prove that AI agents could participate in the sales process for top-tier media events—including both traditional TV and streaming.

"At the time they were all just like, maybe one day, but not today. ... When it comes to the point of transaction, no, we're not there yet. In fact, we're far from it." – Tim Peterson (00:34)

2. Why Start with Live Sports and TV?

[02:21 – 06:59]

• Live sports like the NFL are among the most valuable and complex advertising environments, making automation both challenging and vital.
• Linear (traditional) TV is still 80% of impressions, but much of it isn’t digitized or available programmatically; agentic AI expands the reach beyond what “programmatic” can do.
• The aim is not to replace programmatic, but provide more options and greater operational efficiency across linear, streaming, and hybrid deals.

"We're able to use the technology to show the potential...to automate the full TV buying process, all encompassing." – Ryan McConville (03:13)

3. Technical Implementation: Seller and Buyer Agents

[06:59 – 14:12]

• Seller agents were created for both NBCU’s linear inventory and Freewheel's digital streaming (Peacock).
• These AI agents accessed proprietary APIs to pull available ad units, forecast impressions, package deals, and even simulate identity-backed reach forecasts—tasks typically performed by humans with spreadsheets and dashboards.
• The system’s core is the MCP (Model Context Protocol) server, allowing agents to interact with the same foundational data as human planners, but rapidly and programmatically.

"AI is the new UI...We're actually not creating something net new under the hood. ...What we're essentially doing is making those APIs available to these MCP servers so that agents can just ask the questions of those applications." – Ryan McConville (09:54)

Memorable Analogy & Clarification

• The MCP server is likened to a website for AI agents: just as a human browser is presented with HTML content, MCP delivers comprehensible data for agents.
• Data cleanliness and robust APIs remain absolutely critical: “The biggest amount of work that we need to do is still in the sort of the old school realm of data cleanliness and data standards.” (Ryan McConville, 14:32)

4. Human Oversight and the Buyer's Role

[15:51 – 21:26]

• Despite automation, humans remain crucial for decision-making and oversight; AI agents act more as advanced productivity tools.
• Media buyers interact with buyer agents (chatbots), speeding up the process, but human approval is still mandatory at key checkpoints.

"You are still very much going to have humans in the loop understanding what they're buying and approving what they're buying—and along every kind of step of this process for the foreseeable future." – Ryan McConville (16:20)

Travel Agent Analogy

• The evolution from manual travel agents, to Kayak-style self-serve, to AI agent “human-like” interaction: “Now we're kind of going back to the travel agent. But now the travel agent is an AI agent.” – Tim Peterson (19:13)

Accountability in Automation

• If campaigns underperform, responsibility remains with the human parties, not the AI agents or protocols.
• “If the campaign doesn't perform, ... the buyers and sellers have a discussion.” – Ryan McConville (21:53)

5. Executing the Proof of Concept—NFL Playoff Test

[22:53 – 28:31]

• Addressing the challenge of trust: Big-ticket, live-event transactions carry stakes similar to high-value travel bookings, so buyers need assurance the process is secure and accurate.
• Future needs: Industry standards, trusted registries, and protocols (like ads.txt) will be vital for confidence and authentication of agents.

“How do you know that your agent is talking to the real NBCU agent? … I think all that stuff is going to take time and work, you know, to create, to create trust.” – Ryan McConville (24:10)

Outcome of the Test

• The proof of concept worked: The agentic buying process completed an actual campaign transaction for a live NFL playoff game.

“It works—it was a functioning technical proof of concept that accurately represents what the buyer wants to buy and what the seller has to sell.” – Ryan McConville (28:02)

6. Where Will This Lead? Buyer Trust & Technology’s Evolution

[28:43 – 31:04]

• Full trust in agent-to-agent, end-to-end automation will grow as standards and comfort mature—mirroring the broader pattern of internet adoption.
• “A lot of people want to paint the AI revolution as something that's never happened before. ... I think there's a lot of evidence that there's analogies to the past, and we've talked about some of them.” – Ryan McConville (29:36)

7. Impact on the End Viewer—The Next Level of Relevance

[31:04 – 37:53]

• Agentic AI’s potential goes beyond workflow; it radically increases the contextual relevance and personalization of ads.
• With large language models, NBCU can identify hyper-specific scenes across its vast library, enabling precise placement—for example, “family scenes around campfires” for a camping brand.
• In live sports, real-time signals (like a fumble) can trigger creative that's contextually matched to game moments.

“We can load all of that into a search function and it will scan 30,000 assets ... and it will return in seconds the scenes that are the most relevant and also suggest the ad pods where you would place that ad.” – Ryan McConville (33:04)

The “Targeting Trilogy”

• The evolution: the right ad, to the right person, at the right moment.

"The recall of the ad, the website visits, all of those metrics ... they all go up. So we know that this sort of trilogy works really well..." – Ryan McConville (35:43)

• Oreos “dunk in the dark” moment is cited as a precursor to what’s possible at scale and in real-time with this tech.

“Imagine that at scale and automated. ... The ability to do real-time dynamic creative optimization and actually build the creatives kind of on the fly.” – Ryan McConville (37:17)

Notable Quotes & Memorable Moments

• “AI is the new UI.”
  – Ryan McConville [09:54]

• “The biggest amount of work that we need to do is still in the sort of the old school realm of data cleanliness and data standards.”
  – Ryan McConville [14:32]

• “These are productivity tools that speed up access to information ... but you are still very much going to have humans in the loop.”
  – Ryan McConville [16:20]

• “How do you know that your agent is talking to the real NBCU agent?... trusted registries... things like ads.txt...”
  – Ryan McConville [24:10]

• “It works—it was a functioning technical proof of concept that accurately represents what the buyer wants to buy and what the seller has to sell.”
  – Ryan McConville [28:02]

• “I've been calling it the targeting trilogy. So the right ad to the right person at the right time.”
  – Ryan McConville [35:43]

Timestamps for Important Segments

• 00:34 — Industry skepticism about AI agents in ad buying
• 03:13 — NBCU’s rationale for starting with live TV & sports
• 09:54 — “AI is the new UI”: technical transformation explained
• 14:32 — The importance of data cleanliness in effective automation
• 16:20 — Human oversight and productivity, not replacement
• 19:13 — Evolution of ad buying: travel agent analogy
• 21:53 — Accountability: Where does the buck stop?
• 24:10 — Building trust: authentication, industry standards
• 28:02 — Proof of concept outcome: Agentic ad buy for NFL game worked
• 33:04 — Hyper-contextual ad placement (“campfire” example)
• 35:43 — “Targeting trilogy” and rise in performance metrics
• 37:17 — Real-time dynamic creative at scale

Tone and Takeaways

The discussion is candid, forward-looking, and at times playfully skeptical (“How crazy were you all for doing this?”). The tone remains both practical and optimistic, with an emphasis on the tangible and near-term impacts, not just distant AI hype. There’s clear excitement about agency and advertiser efficiency gains, but persistent reminders of human oversight and incremental trust-building.

Bottom Line:

NBCUniversal’s experiment demonstrated that AI agents can automate complex ad transactions even for premium, high-stakes TV inventory. This unlocks not just workflow savings, but a future of radically more relevant, contextually tailored ads for viewers—while human expertise and oversight remain essential.