Yes, AI agents can buy live sports advertising, but only part of the buy and only with people signing off. Agents can gather availability, assemble a package and negotiate the deal before the event, while the ads inside the live broadcast are still placed by the seller's ad systems. Exceptions such as exclusivity conflicts, creative that fails review and make-goods still go to a person.

TL;DR. NBCUniversal tested AI agents on a buy against a live NFL playoff game in January 2026, with a human at either end of the agent conversation. Live sports is harder to automate than on-demand video because start times move, breaks are unpredictable and audiences spike, which is why IAB Tech Lab is writing a Live Event Ad Playbook for forecasts and concurrency signals. The terms that matter most, make-good triggers, category exclusivity and creative approval, need fixing before kick-off and recording where both sides can see them.

Can AI agents buy live sports advertising today?

AI agents can buy live sports advertising today in supervised tests rather than as routine practice. In January 2026, NBCUniversal, the agency RPA, Newton Research and FreeWheel used AI agents to buy an ad against a live NFL playoff game, Digiday reported. "It works. It is a functioning technical proof-of-concept that accurately represents what the buyer wants to buy and what the seller has to sell," NBCUniversal chief product officer Ryan McConville told the publication.

The agents in that test gathered inventory availability and impression forecasts and packaged a proposal for the buy-side agent, and a human still sat at either end of the conversation. McConville was clear about the distance still to travel: "We are a ways away from having this fully productionalized where multiple agencies are using this day in and day out to replace current workflows," he said.

Why is live sports harder for an AI agent to buy than on-demand video?

Live sports breaks the assumptions most automated buying makes about time and volume. IAB Tech Lab's Live Event Ad Playbook says live event inventory presents "a unique set of operational and technical challenges", driven by "the volatility of live event timelines" and the need to deliver ads to millions of viewers at once. A match can go to extra time, a weather delay can push every break back, and the exact number of breaks is unknown when the deal is signed.

Concurrency is the second problem. Index Exchange chief executive Andrew Casale described the challenge streamed live sports poses for the programmatic supply chain as a "tsunami of supply" in January 2025, according to Digiday. The same report noted that Netflix's Christmas Day NFL games averaged 24 million viewers, so selling that audience impression by impression would mean handling 24 million bid requests within milliseconds.

Buyers feel concurrency as uneven delivery. When large audiences tune in at the same moment, buyers cannot control when their ads serve or how much budget goes into high-demand auctions, AdExchanger reported in April 2026. StackAdapt's live sports workflow, covered in that report, added controls for frequency, budget pacing and filtering inventory by daypart or league.

The third problem is that most premium video is not bought in an open auction. The Trade Desk chief executive Jeff Green said in January 2025 that 75% of CTV ads are sold through direct deals, Digiday reported, and direct deals are where a negotiating agent has the most to do.

What part of a live sports buy can an AI agent handle?

An AI agent's useful work in live sports sits mostly before and after the event. Before the event, an agent can read schedules and forecasts, compare packages, negotiate a deal and draft its terms. During the event, ad decisions happen in the seller's systems at a speed no negotiation could follow.

IAB Tech Lab's playbook draws a similar line. Because live events are high-value inventory often sold through deals, the Deals API comes into play first, and OpenRTB "executes the transactions when the event actually occurs", according to the LEAP page. Its Forecasting API, in public comment until 20 March 2026, lets broadcasters publish event schedules, expected ad breaks and audience forecasts, while the finalised Concurrent Streams API shares near-real-time viewer counts.

How much an agent is trusted to do alone is a separate question from what it can technically do. Alkimi, an enterprise platform for agentic advertising, describes this with a five-step scale it calls earned autonomy: Observe, Recommend, Draft, Human-approved action and Bounded automatic action. The NFL playoff test sits at human-approved action, where agents prepare and negotiate the order and people approve it.

How do make-goods work in a live sports deal an agent negotiated?

A make-good is what a seller owes when it misses a delivery commitment, and live sports makes misses more likely. Television has long treated this as routine: make-goods, also called audience deficiency units, are planned for as part of the business, Digiday reported in 2021. The same report described networks seeking "fluidity" deals that let ads run across linear channels and equivalent streaming programming, partly to control under-delivery.

In an agent-negotiated deal, a make-good only works if its trigger was fixed at agreement. A usable clause names the metric, the party that measures it, the threshold and the window, such as a single match, a tournament or the whole flight. Live events also need clauses that on-demand deals rarely meet: a postponed or abandoned match, a game that overruns and shifts its breaks, or a broadcast moved to another channel or stream.

In a hypothetical case, a buying agent books a package across eight matches of a tournament with an audience guarantee, and one match is abandoned at half time. If the deal says whether that match counts towards the guarantee and where substitute inventory may come from, the agents can process the make-good inside their mandates; if it says nothing, people end up arguing about it after the money has moved.

The mandate should draw the line clearly. An agent can reasonably accept like-for-like inventory in the same competition, while credits, refunds and substitution into other programming go to a person. How do make-goods work when an AI agent negotiated the deal? covers triggers and remedies in more detail.

Which approvals should stay with people in a live sports buy?

Category exclusivity is the clearest approval to keep with people. Brands place high value on exclusivity around major events, and a title sponsor such as an airline may demand separation not only from rival airlines but from any ad that shows an aeroplane, McConville told AdExchanger. An agent can check a rule like that, but a person should approve the rule itself.

Creative review is the second. During the 2026 Winter Olympics, NBCUniversal used AI to scan roughly 6,000 ad creatives against brand safety, exclusivity and category protection requirements, AdExchanger reported. Creatives that scored 100% were cleared for delivery automatically, while "questionable" ones went to manual human review every hour throughout the Games.

That split, software for the obvious cases and people for the doubtful ones, is a workable template for agent-bought live sports. Creative also has to be approved before the event: Disney's Jamie Power listed "the ability to preingest creative and have it ready to go" among the capabilities its live certification for ad tech platforms checks, Digiday reported.

Four further decisions belong with people:

committing budget above the threshold set in the mandate;
agreeing exclusivity and separation terms;
accepting any make-good other than like-for-like inventory;
approving changes on the day, such as a fixture moved to a different slot or stream.

Human approval in an agent-negotiated campaign sets out what sign-off means in practice.

Where do standards fit when agents buy live sports?

Two layers of standards meet in a live sports buy. The agent-to-agent layer is led by IAB Tech Lab's AAMP, the Agentic Advertising Management Protocols, whose OpenProposal (AAMP 3.0) draft is open for public comment until 10 November 2026. AdCP, the Ad Context Protocol published by AgenticAdvertising.org, describes itself as the open protocol that lets AI agents discover, plan, buy, sell and measure media.

The delivery layer for live events is LEAP, the IAB Tech Lab playbook. Its aim, in Index Exchange lead product manager Catherine Cho's words in the August 2025 launch announcement, is to "tame the chaos with standards and make every streaming impression opportunity biddable and addressable by buyers".

Frequently asked questions

Can an AI agent bid on live sports ads in real time?

Ad decisions inside a live stream are made in milliseconds by the seller's ad server and programmatic systems, not by negotiating agents. An agent's role is to set up the deal, budget and pacing rules before the event and to review delivery afterwards.

What happens to an agent-bought ad if the match is postponed?

That depends on what the deal said at agreement. A well-drafted live sports deal states whether a postponed match is rescheduled, replaced with like-for-like inventory or credited, and who approves each option.

Do AI agents replace sports sponsorship negotiations?

Not on the evidence so far. McConville told Digiday that OpenRTB does not really handle bespoke formats or sponsorship integrations, and that agents could start to automate that higher-end part of the market, which still involves people agreeing terms.

Who is accountable if an agent-bought live sports campaign under-delivers?

The seller remains accountable for its delivery commitments, as with any guaranteed deal. The buyer is accountable for the mandate its agent worked to, including the make-good terms it allowed the agent to accept.

Can smaller advertisers buy live sports through automated channels?

Programmatic access to live sports has widened, and StackAdapt made a live sports campaign workflow generally available in April 2026, according to AdExchanger. "You don't have to be an enterprise-level company to run ads during the Olympics," Silverback Strategies' Sarah VanLandingham told the publication.

What does this mean for buyers of live sports advertising?

AI agents can already prepare, negotiate and book parts of a live sports buy, while the event itself is run by the seller's systems and the exceptions by people. The deals that hold up are the ones where make-good triggers, exclusivity rules and approval rights were written down before kick-off and kept where both sides can check them.

Alkimi is built around that shared record. Its Marketplace works within IAB Tech Lab's AAMP framework today, is being built to be AdCP compatible and keeps human approval rights at the deal level, with each agreement written into a DealSheet that both parties hold and can retrieve independently. In its Agree. Transact. Verify. simulation, 95.3% of deals had mismatched records when each side kept its own, against 0.19% with a shared record.