AI and Adtech in Operation
I've moderated several panels recently where we’ve discussed AI in advertising, but some conversations have seemed more theoretical than operational. 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.
“At Magnite we're embracing AI. There are a number of ways that plays through our offerings," says Rory Edwards, VP, Product Demand Solutions at Magnite. The first, he says, is traffic shaping. “How do we take one and a half trillion ad requests that are coming into our platforms—knowing that not all of those are going to be matched with ads—and find the ones that are most likely to match to an advertising campaign and help those to transact?”
The approaches he describes in the following sections are helping to answer this question.
Moving Definitions
Structured information about content is not always provided by the seller for numbers of reasons. Sometimes they don’t have the ability to provide the data. At other times they might want to provide the data but other entities hold the rights to it.
“As of today, [providing structured data] is definitely not standardized across all of the different CTV sellers,” says Edwards. “Some may provide cleaner signals—this is news, this is sports—others may not. When you’re trying to make a market where advertisers want to buy across many sellers, that can be a challenge where they say, ‘I want to target sports, but there are 150 different signals and they change every day.’”
One area that remains challenging throughout the industry, he says, is contextual targeting. “The buy side wants to be able to target content very precisely, and the sell side maybe has different opinions,” he explains. In instances where everyone is okay with that, we do have the problem that there’s no standard way for how content signals are being provided, which makes it hard for that targeting to happen.”
Magnite is addressing problem, he says, by investing “in tools that use large language models to look at all of the ad requests that are coming through from our publishers.” When publishers do provide content signals, Magnite aims “to normalize that and create a way that those can be standardized.”
This makes it possible to identify 50 categories that can be made available to buyers so they can target certain genres with more consistency without having to try and get everyone who’s selling supply to adopt the same taxonomy.
Advertising Auto Correct?
Magnite is also paying significant attention to ad quality. “With more and more advertisers moving into real-time bidding, programmatic-type transactions,” Edwards says, “sellers want to tap into as much of that demand as they can. But when you start moving from 100 advertisers to thousands, protecting that consumer experience is still very important.”
How do sellers monitor the quality and content of the ads and make sure it adheres to their guardrails? “That’s an area where we’ve employed AI to do automatic content review of ads and provide tools to sellers,” he explains. “We call it Ad Quality. It’s a way to consume all of the ads that are coming through, write business rules, and have the AI do the classification so that your ad ops team doesn’t have to spend their entire day attempting to identify any new creative and apply those policies themselves.”
Consumer Behavior
Magnite also leverages AI to do classification where they can create cohorts or audiences based on behavior or affinity for certain types of content. “That is a very big investment for Magnite,” Edwards says. “We sit in a pretty good place because we work with the vast majority of folks in both the CTV ecosystem and the traditional web and mobile.”
Magnite can take all of that insight to create audiences to both inform the seller of their audience, but then also help the buy side to find certain types of consumers across many different sellers.
Intelligent Pricing
Edwards says Magnite is helping publishers to define intelligent price floors for each ad opportunity to help find the right price point in the marketplace given all those shifting supply and demand trends. “This is a machine learning-based model that the publishers we work with can choose to employ to more dynamically adjust to market dynamics,” he explains. “You can really get more out of things on a day-by-day basis by leveraging the AI-driven pricing mechanics," he says.
Increasing CPMs is only part of the purpose. More immediately, he says, the issue is, “How do you find the right price point that’s going to help that one ad opportunity transact? Sometimes it might be higher, sometimes it might be lower.”
With this approach, he says, Magnite has seen customers achieve a 5% increase in monetization.
Media Plan Agent
The challenge for any media plan is finding ways to most effectively accomplish the goals for a campaign: who to reach, where to reach them, what to spend.
This approach, Edwards says, has traditionally been “very manual because I need to go look through all the audiences and figure out what audience segments might align to my target customer. I need to figure out where I can find those audiences and what price I should pay.”
But what if you could work with a ChatGPT type of interface to identify how to fulfill a media plan? These tools are going to give recommendations but still allow media planners to refine and make decisions on where they do want to deploy their budget, and which partners they do want to work with.
“That’s where AI can bring a lot of efficiency, because even if it won’t do everything for you, it can be your assistant to help you figure out what choices you could make,” Edwards says. “I used to need data analysts who would spend a few hours” working with and summarizing the data, “and then I'd have to review that and then come up with an idea. By bringing AI into that process, you could really condense that timeframe and also reduce the amount of people and headcount you would need to make some of those decisions.”
AI is going to make it much easier to answer two key questions: what were the results, and therefore, what should I do to get more of those results?
Business Rules
Ad Quality is an AI-based classification that allows for business rules to be properly enforced for things like competitive separation or frequency capping. Ad Quality sees an ad come through, evaluates the content in that ad, classifies it, and applies categories to the business and other elements.
Ad quality, Edwards says, can answer classification questions like, “’Is this an alcohol ad or a political ad? What industries is it for? What advertiser is it for?’ That provides the structured data so when a seller wants to say, ‘Here are my rules for how I fill my pods and the things I do or don't want to do,’ they have all of the structured data they need.”
This tool, he says, has achieved 95% plus accuracy.
Interoperable AI
Magnite is currently working with an agency holding company which is building their own agentic AI agent experience that needs to interact with Magnite. “Today, we might say we have a bunch of APIs you can use to do things in your platform. In the future, our agent needs to interact with your agent,” Edwards says. He sees a common framework developing for agents to interact.
And that, in brief, is some of what Magnite is doing to work with AI tools which will change how advertising is planned, bought. and sold.
Related Articles
Move builds on momentum of Magnite's Home Screen and Tiles ad formats to meet growing demand for performance-driven ad experiences
26 Aug 2025
In this Q&A with Crystal Foote, founder and head of partnerships at Digital Culture Group, Foote discusses Audience Resonance Index (ARI), "the first AI-platform engineered to predict not just who will engage—but why." We explore how ARI benefits from AI, what ARI measures, what brands ARI is positioned to serve, how ARI safeguards privacy, and more.
19 Aug 2025
What are the key strategies, partnerships, and tech stack elements that make biddable advertising profitable in FAST and other ad-supported CTV platforms in 2025? Fremantle's Laura Florence, Future Today's Tim Ware, and Chris Pfaff Tech Media's Chris Pfaff break down how each of their companies approach biddable CTV sales in a conversation at the latest Streaming Media Connect 2025.
30 May 2025
The buzz around contextual targeting is ramping up in the online advertising world, with the promise that it will solve issues surrounding privacy, transparency, and more. But where is its impact registering in the ad-supported streaming ecosystem? TVREV's Alan Wolk, Fremantle's Laura Florence, Philo's Aulden Kaye Yi, Roku's Charlie Goodman, and Vevo's Melissa Sofo offer their perspectives in a clip from the latest Streaming Media Connect 2025.
05 May 2025