On Monday, Fortune provided a profile and brief interview of Google’s Vidhya Srinivasan, VP/GM for advertising and commerce. Her remit spans $67 billion in revenue including “Google Shopping and advertising products across Google Search, maps, YouTube, and more.”

From Cannes last week, like many in the industry, Srinivasan emphasized the “creative” element inspired by AI: “We’ve always wanted to serve the right ad to the right person at the right time… Now you can come up with a perfect creative at the moment for the person, because the AI is able to generate things on the fly… It’s changing the game on what is possible in any given moment.” Read more.

Nick Fox, SVP of Knowledge & Information at Google, (Mr. Fox is known for the recent buyout memo) said of Srinivasan and this Fortune article on Linkedin ,

“What a difference a year makes.  Last year at Cannes Lions, Google’s VP / GM of Ads & Commerce Vidhya Srinivasan said she heard mostly curiosity (and even some trepidation!) about AI in advertising.  This year, that curiosity has become confidence.”

“This shift is in large part due to the work Vidhya and her team have done over the past year to listen to advertisers’ and merchants’ needs and help them use AI to create more relevant, helpful, and effective ways to connect with users.”

Agent-to-agent AI

In an interview recorded last week and titled, “Why LLMs are Fuelling a ‘Big Bang’ in Ad Tech,” Gurman Hundal, Co-Founder and Global CEO, MiQ briefly discusses the broader implications of AI in advertising as well as his company’s new AI-powered, programmatic buying platform Sigma.

Readily admitting that artificial intelligence and machine learning have been around a long time in advertising, Mr. Hundal notes the rise of large language models (LLMs) as inspiration for this latest wave.

Looking ahead, he added:

“The area I’m really excited about is how we start to make Sigma more interoperable with some of our clients’ own technology developments. And that might not be a front-end login. It might be more back-end types of integrations where they’re connecting to certain parts of Sigma. I think some of this agent-to-agent AI, or their agents-to-our-models, and that sort of thing, will become the next wave of conversations and development that people work on.”

See a bit more on Videoweek.

podcast: TikTok in AI bind

On the AdExchanger Talks podcast, David Kaufman, who leads TiKTok’s Product Development, Product Marketing, and Product Operations departments for monetization spoke about his company’s recent AI and advertising updates with AdExchanger’s Allison Schiff at Cannes last week. Kaufman is a former Google-er.

Hear it on Apple Podcasts.

Even with the seemingly, precarious nature of the TikTok app in the United States given a possible U.S. governmental – albeit delayed – ban, Mr. Kaufman claimed his company and its clients have become accustomed to the hovering cloud and pointed to the app’s continued growth.

Also, Kaufman distanced his company’s “AI + ads” approach from that of Google’s Performance Max (PMax) and Meta’s Advantage +, where each increasingly promise to take the marketer’s budget and then let AI do the rest.

Responding to the “late 2026” deadline posed by Meta CEO Mark Zuckerberg about everything being AI-automated for Meta’s buying platform, Kaufman said [23:39]:

“I think my response is maybe someday, but it doesn’t feel close to me at all. I look at the capabilities of generative AI, and AI in general, and I really see them as tools that help drive marketers to be even more successful. So if you think about what makes an advertisement successful – understanding your audience, being creative, telling a compelling story, introducing a product, all of those things – AI can help you with them, but I don’t think it can replace the creative process.”

TikTok appears to be in a bind here when it comes to its AI plans and may be why they’ve been stressing their creative AI capabilities – though Kaufman added that there will be innovations related to “the auction itself.”

If the AI-led buying platforms such as PMax and Advantage + prove to be the way to go for efficient media buying, the perception of giving over to a similar product from Chinese-based TikTok’s AI may be a bridge too far for U.S. lawmakers.

This AI hesitation could inhibit innovation and revenue at TikTok.

Of course, TikTok could continue to innovate, in parallel, in non-US markets.

Google AI chips advantage

In a Google blog post on Monday, “Ask a Techspert: What is Inference?” sheds light on an artificial intelligence concept (inference) at Google -and elsewhere for that matter. But, the post also spoke to a critical, potential advantage for the company: it’s own chips.

Read the Google “Keyword” blog featuring Google engineer Fenghui Zhang.

Zhang explained:

“Inference in general is the way we actually use the model to do something useful. First, we have to train the model: An AI model will contain the model parameters, model architecture and configuration, which is the code that it needs to execute tasks — and these things combine to carry out the functionality. So inference is what allows us to actually take all that and use it.”

Towards the end of the interview, Zhang brought up Google’s specialized chips known as Ironwood and their importance in the AI and inference discussion (see this April Google blog post about “Ironwood”).

MediaPost’s Laurie Sullivan distilled the Ironwood implications:

“Advertisers will soon have faster hardware and software support to run all types of features being built in artificial intelligence. Google Ironwood, the company’s seventh-generation Tensor Processing Unit (TPU), and the first designed for large-scale AI inference, will launch later this year. The AI accelerator chip optimizes hardware and software for advanced AI workloads, including those in Gemini models.”

Prompt: Deterministic versus probabilistic measurement in advertising… which matters most in five years?

Response from Anthropic’s Claude:

The advertising industry is moving decisively toward probabilistic measurement, and this trend will only accelerate over the next five years.

Why Probabilistic Will Dominate:

The fundamental shift away from deterministic tracking is already well underway. Apple’s iOS privacy changes, Google’s phase-out of third-party cookies, and expanding privacy regulations like GDPR and CCPA have made deterministic measurement increasingly unreliable. By 2030, the deterministic signals that advertisers have relied on for decades will be largely unavailable.

Probabilistic measurement, while less precise at the individual level, offers several crucial advantages for the future landscape. It can operate within privacy-first frameworks, scales across different platforms and devices, and provides insights that comply with evolving regulations. Advanced statistical modeling and machine learning are making probabilistic methods increasingly sophisticated and actionable…

Read more on tipsheet.

---

More stuff

• Opinion: AI Is Killing Marketing As We Know It — So What Comes Next? by Steven Wolfe Pereira (June 24) – Forbes

• Why Does Every Commercial for A.I. Think You’re a Moron? (June 25) – The New York Times (subscription)

• Advertising Could be AI’s Next Frontier. Get Ready for Ads in Chatbots. (June 18) – Barron’s (subscription)

• The ad industry is obsessed with AI — but what about its energy use and sustainability? (June 25) – Digiday (subscription)