Media buying just crossed a threshold: on a Q1 2026 live NFL playoff campaign, NBCUniversal, independent agency RPA, FreeWheel, and ad analytics firm Newton Research ran an entire premium video buy, from planning through optimization, with AI agents transacting directly with each other across linear and digital inventory.
The four companies announced the partnership this week, describing it as the first time AI agents have automated the buying and selling of live sports inventory on linear television. The deal matters less as a single campaign than as a template: it shows what happens when the sell side and the buy side of a media transaction both deploy agents built on a shared protocol, instead of humans trading spreadsheets and insertion orders across a negotiation table.
How the agents actually transact
NBCUniversal and FreeWheel deployed AI sales agents on the sell side, covering linear inventory and digital inventory respectively. Newton Research designed and implemented the buy-side agents for RPA, the agency executing the media buy on behalf of its client. The two sides interoperate using Model Context Protocol (MCP), the open standard that lets AI agents built by different companies discover and call each other’s tools without custom integration work for every pairing.
Newton Research CEO John Hoctor said the agents “interoperate and collaborate with other agents, data and technology,” which is the detail that separates this from earlier automation. Programmatic buying already lets software execute a bid in milliseconds. What is new here is that the negotiation, inventory discovery, and optimization steps that used to require a planner on each side are now handled by agents that can query each other directly, across a linear-digital boundary that has resisted full automation for years.
Why live sports was the test case
Live sports inventory is the hardest place to prove this works. It is scarce, time-sensitive, and has historically been sold through upfront commitments and manual scatter-market negotiations precisely because algorithmic buying struggled with its unpredictability. Mark Marshall, NBCUniversal’s chairman of global advertising and partnerships, said the move will “redefine how inventory is bought and sold,” calling live sports the starting point rather than the destination.
FreeWheel general manager Mark McKee framed it as “a fundamental step” toward what the four companies describe as automated, outcome-oriented buying and selling, where the transaction optimizes toward a business result instead of a fixed rate card.
What agentic buying means for the marketing leader
For a CMO or media director, the near-term impact is less about replacing planners and more about what becomes possible once agents can transact directly. RPA CEO Jim Helberg called it a way to “hyper-streamline strategic media intelligence and transactions in service of business outcomes,” which is agency language for compressing the time between deciding to shift budget and having that budget actually reallocated across platforms.
That compression is the real story. A buy that once took a planner days to renegotiate across linear and digital inventory, pulling avails, checking rates, and re-forecasting reach, can now run as a live optimization loop. The martech stack implication is direct: agencies and brands that have already invested in agentic, audience-ready infrastructure, the same shift behind publisher data monetization going agentic, are positioned to plug into agent-to-agent buying first, because they already have the audience and measurement layers agents need to query.
It also raises the bar on standards work already underway. IAB Tech Lab’s AAMP 2.3 update exists precisely because enterprises need governance rules before they hand transaction authority to agents, and DSPs like StackAdapt have been moving planning into conversational interfaces for the same reason: the industry is converging on agents as the default interface for media buying, not just an experiment.
The choice of MCP specifically matters. Rather than NBCUniversal, FreeWheel, and Newton Research each building custom point-to-point integrations, all three built to the same open protocol, the way websites build to HTTP instead of negotiating a private format with every server. That is what makes this a template other publishers and DSPs can copy without a direct partnership with Newton Research first, and it is why the deal reads as an infrastructure moment rather than a single client win.
The case for caution
Enterprise buyers are not fully sold yet. In an IAB survey of 205 US ad buyers conducted between November 2025 and January 2026, 40% named understanding agentic ad buying and campaign execution as a top concern for 2026, according to eMarketer’s analysis of the data. That anxiety sits alongside the excitement: agencies want the speed agents provide but are still building the oversight layer to catch a bad reallocation before it compounds across hundreds of campaigns.
None of the four companies addressed error handling or human review checkpoints in the announcement, which is the detail marketing leaders evaluating similar deployments should push their own agency and ad tech partners to specify before scaling beyond a pilot campaign.
What to do next
Marketing leaders do not need to build agent-to-agent buying capability themselves to be affected by it. Three moves matter now: confirm which of your DSPs and sell-side partners already support MCP or an equivalent agent protocol, ask your agency what human checkpoints exist before an agent reallocates spend across channels, and treat this partnership as a preview of your next linear and CTV upfront negotiation rather than a one-off pilot. The agencies moving first on agentic buying are the ones that will set the terms for how it works once it reaches the rest of the market.
Source: NBCUniversal
