In the lead-up to Meta’s Performance Marketing Summit 2025 this week, Mobile Dev Memo analyst and podcaster Eric Seufert spoke with Matt Steiner, VP of Monetization Infrastructure, Ranking & AI Foundations at Meta, on how AI is being integrated within Meta’s ad platform.
-
Read the auto-transcript here from Apple Podcasts.
The contents of the interview are a veritable banquet of Meta AI ad strategy compared to the WSJ Meta ads “appetizer” on Monday.
Seufert revisits – with the help of a Meta subject matter expert – much of Meta’s ad strategy which is already public (Lattice/Andromeda/GEM) but, perhaps, misunderstood.
Steiner’s role is at the center of the AI advertising “storm” at Meta. He describes his group’s responsibilities:
“This is the set of teams that build the infrastructure that underlies our ad system, including the ad serving system, the conversion tracking system, the storage systems that house our ads, the batch processing systems that compute data about ads or metrics about how ads have been interacted with, and the machine learning model training system for ads machine learning models.”
“It also includes our teams that do AI research on ranking and recommendation research, as well as our teams that do development of machine learning models, as well as accelerating machine learning model training, and really kind of end-to-end optimizing the training and serving of machine learning models that make predictions about which ad to show a person when they log in to one of Meta’s apps…”
And VP Steiner knows his purview inside and out. He’s able to talk about the details of Meta’s “Generative Ads Recommendation Model (GEM)”, for example, without hesitation.
The highlights of the interview are summarized by Seufert here and include ”The evolution and functionality of Meta’s various AI-enabled ranking and selection tools, from Lattice (transfer learning) to Andromeda (ranking) to GEM (sequences)”
More highlights from Steiner…
-
A large team of Meta AI researchers in the AI ad space are focused on driving performance (33:40) which breaks down in two parts: conversions (retrievals, ranking, Advantage+ shopping automation) and the other side of the ROI calculation – how much you have to spend in order to achieve results. This includes generative AI for creative, business messaging and AI agents.
-
In spite of all the AI and automation, Steiner urges marketers that “testing is the secret to success” in order to drive better results.
-
On AI talent or a lack thereof: Delineating between Generative AI and “prior-generation” ranking & recommendations AI, Steiner was optimistic that the human talent to service these discreet types of AI approaches was more plentiful than ever. (From here, acquiring talent still seems like a key part of any AI-related M&A.)
quotable:
-
Eric Seufert yesterday on X, “…the opportunities [that digital advertising as an ‘AI business agent’] present are astounding and very well could significantly grow the digital advertising market materially as the barriers to entry are lowered and new participants spend their first dollars of media.”
HubSpot and AI
“HubSpot is the first CRM to be deeply connected with OpenAI’s ChatGPT… With this new connector, with just a few clicks the 250,000+ HubSpot customers can bring CRM data right into ChatGPT and ask really sophisticated deep research queries using that data.” – Dharmesh Shah, CTO, Hubspot on LinkedIn (June 4)
Snowflake hire
Data cloud infrastructure company Snowflake may be making a deeper play in ads/marketing with the help of artificial intelligence.
First, ads are marketing are well-represented at this week’s Snowflake Summit in San Francisco – see Kargo pitching AI-powered contextual targeting at the event.
Next, ad tech executive Dennis Buchheim, formerly of Yahoo, Microsoft and IAB Tech Lab among other stops, announced on LinkedIn yesterday that he has joined Snowflake “to lead adtech/martech, media, and entertainment GTM.”
Buchheim shares why he joined saying, in part, “I’ve spent most of my ads/media career building platforms and tools to support publisher monetization and scale ad products. Snowflake’s position near the ‘core’ of the data stack makes it different: enhancing this core can rapidly and expansively drive effective and responsible use of data and AI — more so than effecting gradual change in the ‘outer layers’.” Read more on LI.
ecommerce images
“Meta published a case study of our work on fine tuning llama models for product taxonomy inference” (See case study, PDF) – Tobi Lutke, CEO, Shopify on X (June 4)
agencies v. consultants
Taking aim at CMO budgets (and elbowing aside agencies) in the age of AI, EY announced EY Studio+ on Tuesday which the global consultant described as an amalgamation of “the organization’s full spectrum of services with the world’s leading talent in design, sales, marketing and customer experience to drive growth and transformation for clients in an AI-driven era at a global scale.” See press release.
Consulting conglomerates aching for new areas of growth have been making digital marketing pitches for a while now. This latest pitch appears to omit “digital” and replaces with “artificial intelligence.”
For those attending the Cannes Lions ad festival in France later this month, the Wall Street Journal reports that EY will be rolling out Studio+ in a big way (big yacht?).
The WSJ’s Katie Deighton paraphrases new Studio+ leader Laurence Buchanan: “Clients right now may not see the firm as the ‘most creative company in the world, (…) but they know EY has credentials when it comes to ‘financials, compliance and making this work overall for the business.’” Read the WSJ.
targeting case studies
“Today at the IAB Tech Lab Summit, we announced two new case studies highlighting how Mobian’s agentic AI-powered targeting solutions drive outcomes for brands and publishers alike (…) Full case studies attached.” – Mobian on LinkedIn (June 4)
agencies: LMM
On Tuesday, Brian Lesser, CEO of WPP Media (formerly GroupM) revealed that his media buying agency has a Large Marketing Model (LMM not LLM) “built to drive business outcomes in the AI era” called Open Intelligence.
Read the press release on the WPP Media website.
agencies: AI & creative
CEO Brian O’Kelley of Scope3, which positions itself as an “agentic media solutions” provider, announced that he’ll be participating in a June 10 webinar with Camilo La Cruz and Alex Siddall from Omnicom Media Group to talk about how AI is “unlocking creativity.” Registration is here.
pitching native AI OS
AdTheorent founder Anthony Iacovone has launched Veylan which is positioning itself as a “native AI operating system for advertising.” Read the press release and learn more about the pitch. Notably, former LinkedIn executive Penry Price has signed aboard as an advisor and said on LinkedIn, “I was introduced to Veylan and saw the future of advertising with one demo.” Read more from Price.
more stuff
-
CMOs on AI: H&R Block’s Jill Cress (May 2025) – Google
-
Google DeepMind’s CEO Thinks AI Will Make Humans Less Selfish (June 4) – Wired
-
AMC Networks will use AI firm’s tech both to generate marketing images and help pre-visualize shows that have yet to be produced. (June 4) – The Hollywood Reporter
-
MiQ Sigma: AI-powered advertising technology trained on the most connected, global data set (June 4) – MiQ on LinkedIn