Good Luck Trying to Figure Out Just How Fast AI Will Upend Advertising
Yahoo DSP boss on the speed of the agentic economy
It is becoming increasingly challenging to handicap just how much artificial intelligence promises to fully transform the structure and operations of the digital advertising industry. The narratives being delivered are - extreme in both directions.
One end of the spectrum describes a world where agents replace numerous tasks within agencies and ad tech companies - freeing up humans to do lots of strategic work, or just get laid off. These proponents warn that the ad world at large isn’t grasping the required urgency, and moving way too slowly, while predicting that market leaders will embrace the Jack Dorsey/Block strategy - i.e. replace half the workforce with AI overnight.
At the other end of the spectrum, analysts talk about the resilience of humans - or rather, the intractability of human talent. It’s too hard to rip up all of an industry’s processes at once, they argue. And employees aren’t going to help with their own obsolescence.
Big companies barely know what to do with AI anyhow, many say. If anything, it’s just creating more work (WSJ: AI Isn’t Lightening Workloads. It’s Making Them More Intense).. Plus, marketers are always going to be hesitant to let machines run wild with buying media or making ads without human supervision, right?
Well, at least when it comes to programmatic buying and selling, Adam Roodman, GM of Yahoo DSP, is predicting a rapid shift, driven by agentic ads. “I think it’s moving pretty quick,” he said during the latest episode of Next in Media.
“If I think about the first iteration of AI that really came into the programmatic ad ecosystem, that was definitely machine learning, where you saw ad networks evolve into demand-side platforms…That was probably like a five-year sort of like metamorphosis.”
“This one’s going much, much, much quicker. We went from learning what an MCP (Model Context Protocols - the means through which agents connect to one another) was to standing one up within, I think, six months. It was really quick.”
That very much jives with what Kochava CEO Charles Manning told me a few weeks ago on the show.
Agents, ad agentic advertising, and the vast potential of AI were on full display last week at Marketecture Live. As keynote podcast guest Jeff Green, CEO of The Trade Desk, put it: “I don’t think there is an industry in the world that is more conducive to AI than programmatic advertising. We’re looking at 20 million ad impression opportunities every single second. We’re also representing millions and millions of ad campaigns, with lots and lots of creatives, billions of users on the other side, and we have to do this in 10 milliseconds.”
The exuberance was showcased by a packed MCP session, along with a keynote delivered by one-time TED Talker Jeremiah Owyang, General Partner at Blitzscaling Ventures
“AI transforms the decision funnel for marketers,” Owyang said, while describing a future in which all consumers will collaborate with agents, the web will be a ghost town, and commerce and content will come to you.
“This is the biggest change in marketing since the dot-com boom,” he said. “It will be a year or two before it becomes mainstream.”
To be fair, I’ve delivered zero TED Talks, but I think Owyang may need to get outside of the Valley bubble more. First of all, people really don’t like AI. Plus, normal people like shopping, and are seeking connections IRL more than ever. As Julie Rice, Chief Experience Officer at WeightWatchers put it, people like the tactile, such as when shopping for clothes. “You want to touch and feel the fabric,” she said.
As for the business operations side of things, Owyang may need to spend some time in Madison Avenue, where agencies have been going extinct for 25 years. During a panel I moderated with Horizon CEO Bob Lord and Obele Brown-West, president of Colle McVoy’s Next President, both acknowledged the need to reform agency business models and to become far more transparent with clients. And both saw creative and media being forced back together.
Yet both Lord and Brown-West were optimistic about the role of people - and doubted that brands were ready to hand everything over to agents - or Meta for that matter.
“You don’t need to burn it all down,” said Brown-West.
Yet not long after our keynote, news broke that Horizon was laying off 50 people. Is that an isolated case, or the beginning of a wave? To hear Owyang tell it, in Silicon Valley, he’s seeing top software-as-a-service companies generating $524,000 per employee, while very new ‘AI-native companies” are generating $2.2 million per employee, or 10X the average.
“This is the future that we are looking at, and the expectations” he said. While perhaps unrealistic, agency holding company CEOs, along with ad tech founders, will surely see those kinds of numbers and ask, how do we get there?
Roodman thinks that big agencies are going to invest heavily in agents, as will various third parties, along with media sellers. Thus, the company’s “Your Mine and Ours strategy” is designed to not bet on any one side, but instead to ensure that all these new agents work together.
Yet I wonder if this Everyone Will Build Agents theory is reminiscent of when ad agencies bought their own ad networks and ad tech, before eventually realizing it was not in their wheelhouse. A report from Gartner is already warning against brands getting too locked into one agency’s tech platform.
“We don’t think having agent tech is a strategy,” said Roodman. “It’s how you operate it is a strategy.”
Maybe agencies will just get really good at using others’ agents. According to Madison and Wall’s Brian Wieser - Wall Street’s take on agencies has been “mostly incorrect. “I am increasingly positive about agencies. A person plus a machine will be the machine every time.”
In the meantime, it will be fascinating to see how this will play out in the media market. As I wrote about last week, in the TV world, there is still a hesitancy to open up too much inventory to programmatic channels. So will they draw the line at agents? “Our inventory footprint of what we consider the open web is going up, not down,” said Roodman. “Yeah, there are pockets of web inventory that maybe are not trending up, but there is so much new addressable inventory in audio and CTV that I’m not concerned.”
As for the ads themselves, Roodman thinks that the industry may have gotten a false positive when it comes to generative AI, and the idea that creative output will soon be automated and expanded exponentially.
That was somewhat illustrated during a super interesting session last week with Sofia Colucci, Chief Marketing Officer at Molson Coors, who described the company’s internal creative process, an attempt to quantify exactly what specific elements need to be in a video ad in order for it to resonate with consumers (and score high in various ‘neuro testing’ systems the brand employs).
“Creative is about driving the business for us,” she said. “We wanted to be the most effective creative beverage marketer, and we needed common language and shared definitions of creative effectiveness.”
Sounds like the perfect thing to train an AI Creative Agent on! Yet to me, this feels like brands trying to fully apply the scientific method to something that is always going to be subjective, and somewhat random - a great ad.
Indeed in the ultimate “humans still matter” admission, Colucci told the story of how the brand was planning an ad campaign for the beer Blue Moon, which consumers typically drink with an accompanying orange slice.
The beer brand had several creative concepts in mind, when spokesperson and Saturday Night Live cast member Colin Jost suggested a wacky idea featuring anthropomorphic oranges clamoring to be added to a Blue Moon Draft. The team at Molson Coors didn’t think it was a great idea.
It turned out to be their best performing ad.