How to Leverage AI in Ad Insertion Workflows for Live Sports Streams at Scale
The extraordinarily high concurrence of viewership with tentpole live streams complicates every aspect of content production and delivery, and ad insertion is especially hard to scale, according to FreeWheel VP of product architecture Jeff Ellin and IAB Tech Lab EVP of product and COO Shailley Singh in this clip from Streaming Media Connect 2026, in large measure because of the pressure it puts on system responses not to add latency to the stream. Ellin, Singh, and Streaming Media contributing editor Nadine Krefetz discuss both the complexities involved and how AI can improve response times, ensure better decisioning, and prevent systems from getting overwhelmed everywhere in the supply chain when ad requests come in, particularly with programmatic ads.
Why Ad Insertions Get More Complex as the Events Get Larger
Krefetz sets up a scenario to ask about live workflows: “So we have a big sports game going on. We might have, oh, I don’t know, the World Cup, or we might have any other big tentpole event. What are some of the issues that we have [with] AI that we’ll work with? Or a concern that we have when we’re looking at ad insertion for a really large-scale event?”
Ellin notes that large events do have more complexity when it comes to scaling up to handle ad breaks. “Let’s say you’re watching NFL football on Peacock on Sunday night. There’s millions of people watching it. A timeout gets called and immediately goes to commercial. Those ads are hard to scale. I know Shailley [Singh]’s going to talk [about] some of the stuff probably that IAB’s doing, and FreeWheel’s involved with helping set those standards. But to scale that programmatically, there’s so many people in the chain: SSPs, DSPs, ad stitchers, ad management, ad servers. All of them need to be able to handle an ad request that comes in within a very small amount of timeframe,” he says, adding that IAB Tech Lab is working on the relevant “protocols to help everybody in that supply chain to become ready to be able to accept that and be primed to respond in a time needed. And that’s going to be helpful in making sure that in a programmatic sense, live events don’t overwhelm the system and lose an opportunity because if you don’t respond quick enough, the ad doesn’t show.”
What IAB Tech Lab Is Currently Doing
Singh shares that IAB Tech Lab is addressing two key areas in live streaming: latency and metadata. Buyers need to understand a stream at a metadata level in order to make good decisions, “especially in a programmatic transaction,” he notes. “So we’ve been working on multiple things to support this. So the first one is, we have a program called LEAP, Live Events Ad Playbook, that is a bunch of what we call Sitecore APIs to inform everybody in the chain—like Jeff [Ellin] said, there are ad servers, DSPs, SSPs—to inform them about forecasting the audience so that they can prepare their compute capacity accordingly for each live stream so that they’re not short of compute capacity or bandwidth capacity to address incoming ad requests. And then there is the actual live stream API that gives you the actual number of users or viewers that are online.”
The second thing IAB Tech Lab is working on is live attributes that provide information on what a stream is about. For example, is it a live game, or is it a game that was played earlier but is being streamed for the first time? This helps “the buyers to decide how they want to proceed with bidding or buying that and setting up the deals, etc., that needs to be in place,” Singh explains. “We are also working, like I said, on the creative API so that you have a pre-approved, ready-to-serve creative that also adds to improving the latency factor. And then we also introduce something called ARTF, which is like an agent that is living inside the host platform, which can actually compress the time to make decisions by 80%.”
Other IAB Tech Lab Initiatives
Singh discusses two more avenues of live-stream support that IAB Tech Lab has in the early stages of development: transparent bid caching and pre-fetching.
Bid caching will be able to address immediate ad break requests, Singh says. “You can have a bid that is designed for the whole program so that you don’t have to go back and forth and bid again and again, each impression, but you can bid once for the duration of, say, a three-hour NFL game and then you can buy the ads with the same bid. So that itself is going to reduce a lot of the latency.”
Pre-fetching has been in practice for a long time, Singh notes, “but the CTV space is slightly different. The platforms are slightly different. [There’s] a mechanism of pre-fetching ads so that the ad is ready on the device and you don’t have to go back and forth between two different cloud systems and reduce the latency there.”
AI Can’t Function Effectively Without the Proper Context
Krefetz wonders, “Now is AI involved in all of these? Because people have spoken to me about pre-posting for quite a while now.”
“Yeah,” Singh acknowledges. “If you are going to use AI in making decisions in supporting this workflow, AI needs all of the context,” he cautions. “And these are the things like forecasting API, live stream API, the live attributes in RTB, they’re all going to provide the AI system with the right context, which is pretty well known. If the AI systems don’t have the right context, they tend to hallucinate. So all of these together will provide the exact context under which the AI system can actually operate and make decisions.”
Join us 11-13 August 2026 for more thought leadership, actionable insights, and lively debate at Streaming Media Connect 2026! Registration is open!
Related Articles
Globo director of adtech Ana Beaklini says latency has traditionally been the biggest problem the Brazil-based Globo has faced in its efforts to provide seamless ad insertion for high-stakes, high-concurrency, ultra-low-latency live events like the World Cup. In this conversation with Streaming Media contributing editor Nadine Krefetz at Streaming Media Connect 2026, Beaklini explains how Globo has addressed this issue and the best practices the company will implement at massive scale for upcoming World Cup live streams and OTA broadcasts and AI's role in hyper-personalizing Globo's live ad decisioning.
28 May 2026
What is server-guided ad insertion (SGAI) from a technology standpoint; how does it differ from client-side ad insertion (CSAI) and server-side ad insertion (SSAI); how does it impact or change existing streaming workflows, infrastructure, and delivery; and what benefits does it provide to streaming publishers, advertisers, and viewers? Bitmovin's James Varndell and Ring Digital's Brian Ring break down SGAI, its advantages, and a key emerging use case in this discussion from Streaming Media Connect 2026.
16 Apr 2026
With an eye to providing more insight into real-world deployments and the operational side, online adtech firm Magnite volunteered to give me insight into how AI fits into their current operational processes.
05 Sep 2025