AI Briefing: Media Buying Summit highlights judicious adoption of AI ad tools

AI gives answers, but not always the right ones, said Marilois Snowman, founder and CEO of media services agency Mediastruction and SaaS platform FutureSight. Speaking onstage about media mix modeling, Snowman said AI can help extract, transform and load data faster and more efficiently. But she pointed out it’s important to use tools judiciously to make sure the outputs are accurate.

“We’re still thinking more about machine learning algorithms because we can understand how they got to the answer,” Snowman said. “These outputs are only as good as the data that they’re given, so if we want to understand where things went awry we’re following the margin of error of what our models are forecasting and what’s actually happening. And if we’re finding that margin of error is just not improving, with ML models we can better understand why that is.”

In a keynote conversation that kicked off the summit, Kamran Asghar, CEO and co-founder of independent media agency Crossmedia, said clients are eager to test new AI tools for media buying, insights and content creation.

“We use AI in some predictive modeling in order to understand where our next best dollars should be spent once we have evidence that things are working,” Asghar said. “So having it by your side, but not be the end-all be-all is the standard step.”

There’s also the challenge of potential platform bias with AI tools. Asked about the revelations from the Google ad tech antitrust case, Snowman said lost trust creates a problem for measurement: “We have to play in their sandbox — pun intended — but we don’t have to rely on them for measurement,” she said.

Warnings against over-relying on automated ad-buying was a key theme throughout the summit. Tucker Matheson, co-founder of the digital strategy firm Markacy, said the past decade of digital media has created a bias toward platforms. Large advertisers have the resources for media mix modeling but that’s not always the case for smaller brands. Matheson also it might not be wise to anchor media budgets based on what platforms provide in terms of traffic acquisition costs or return on ad spend.

“It’s just so easy to get into Meta and Google and see the data and click buttons,” Matheson said. “Platform metrics still are important, but I think they need to be looked at a little bit differently. Like for us, we all know the proliferation of creative testing on Meta and TikTok. I think those platforms give you good click data to support creative testing.”

(Coincidentally, Google this week also debuted new ad features for Performance Max and search campaigns including new ways to optimize campaigns and track performance. The updates for Google Ads API version 18 include new ways to query placement-specific data and content metrics for display, demand gen and video campaigns.)

Meanwhile, AI’s growing presence in creative ad generation is being felt across the marketing spectrum. Balancing brand voice and accuracy is key for customer-facing uses of AI, said Kevin Rettig, Marriott’s senior director of marketing platforms & privacy. In a conversation about personalization, Rettig mentioned Marriott hasn’t done much yet with generative AI ads. However, it is experimenting through a recently rolled out virtual assistant powered by ChatGPT and open-source outlets. Another challenge is maintaining brand voice across the hotel giant’s dozens of brands that each have their own brand voice.

“When you think about AI and content generation, each brand is very protective of what their brand voice is,” Rettig said. “We have to balance that with the automation and someone else writing or doing the creative piece. I’m not saying we wouldn’t do it and I think all brands have that sensitivity, but with 30 brands that really feel their brand voice is unique, we have to be really mindful of that too.”

Outside of AI, the hotel giant is exploring plenty of ways to personalize marketing through its Bonvoy loyalty program and other data sources, said Rettig. He noted the company uses first-party data to understand customer hotel preferences, power targeted promotions and enable ads bought through the Marriott Media Network. While Marriott is exploring ways to expand its data pool through third-party data sources and clean room partnerships, it also uses data verification tools like Neutronian and Truthset.

Prompts and Products — AI news and announcements 

• X is updating its user terms and privacy policy for users, which include changes related to how it collects, shares and uses data for training X’s AI models. The new terms, which go into effect in November, also allow X to share user data with “third-party collaborators” for “independent purposes” including training AI models, “whether generative or otherwise.”
• Google announced new AI shopping features for personalization ahead of the key shopping season. It also announced Prabhakar Raghavan, the company’s top exec for search and ads, is changing his role to become chief technologist working directly with CEO Sundar Pichai. Replacing Raghavan is Google vp of product management Nick Fox. The changes are largely the result of Google moving the Gemini app team under Google Deepmind.
• Amazon’s annual UnBoxed event featured new machine learning tools for creating campaigns, finding related audience segments, and monitoring various metrics. The e-commerce giant also announced a new AI creative studio and audio generator to help scale campaigns. 
• Adobe debuted new AI creative tools at its annual Max conference, including a new GenStudio for performance marketing along with a related partnership with Microsoft Advertising. Adobe also rolled out its new AI video model, which the company says is trained on licensed content.
• Taboola introduced a new generative AI ad assistant called Abby that lets marketers create and manage digital campaigns through a chat-based interface.
• Perplexity announced its first version of Enterprise Search called “Internal Knowledge Search,” which will help employees do research at work using LLMs.
• Mozilla and the Open Markets Institute partnered on a new report about AI and competition.

AI stories from across Digiday

• The AI revolution is creating more demand for AI librarians. (Digiday)
• Publishers are receiving credits from AI companies to use their large language models via recently signed content licensing deals. (Digiday)
• Marketers aren’t won over by new AI-powered search tools. (Digiday)
• As TikTok rolls out Smart+, marketers grapple with AI’s advertising puzzle. (Digiday)
• Agency execs balance automation and strategy with new ad formats. (Digiday)

AI news from elsewhere

• Security researchers found a way to use malicious GenAI prompts to collect personal data based on chatbots users’ past conversations. (Wired)
• YouTube took a step toward labeling videos using C2PA standards. (The Verge)
• The British government said new legislation might be needed to address uncertainty related to AI and copyright. (Lexology)
• Generative AI is creating a flood of fake AI voices that sound like Kamala Harris and Donald Trump (The Washington Post)
• Dane Stuckey, Palantir’s former chief information security officer, has joined OpenAI as its new CISO. (TechCrunch)