GOOGLE CLOUD NEXT, LAS VEGAS – It's been a busy few days and we've had more meetings here at Mandalay Bay than we can count (or, let's be honest, even remember). While not every one will make it into a full fledged story, many of our conversations yielded some worthwhile quotes that got the wheels turning upstairs.
Without further ado, here are some of the most interesting bits and bytes from our experience here at Google Cloud Next. Oh, and don't miss our highlights reel above.
Intuit's 4 million models
"That's how you get to 4 million models." — Intuit SVP and Chief Data Officer Ashok Srivastava, explaining why the company is running so many AI models internally.
Yes, you read that right. Intuit is running not 400, 4,000 or even 400,000 AI models, but 4 million to help power user experiences like transaction categorization.
"We would need to have a set of models that are built on a population of Quickbooks users and, then second, for a given user, we'd need another set of models. So when you add that all up, it turns into 4 million models," he said. "This is customer-level personalization."
Of course, our natural next question was how on earth Intuit manages 4 million models. Srivastava's answer? "A very powerful monitoring system that we've built."
Google Cloud on network autonomy
"It’s just chaos when it comes to trying to bring it all into a foundation upon which you can get started." — Angelo Libertucci, Google Cloud's Global Head of Industry - Telecom, on how issues with data are hindering efforts to make the IP core fully autonomous.
Libertucci told Fierce that Google is working with operators to bring full automation to wireless networks first, in part because 3GPP standards mean that operator data is a bit more organized. That's not the case on the IP core side, he said.
There, operators need to tackle structured and unstructured data in a smattering if different formats, making the task of building a data foundation on which AI can run that much harder.
AI helping avoid the hire-and-fire cycle
"I look at this as protecting the team we have right now with the utmost job security. You know we've never laid anyone off. You know we're never going to." — Jason Bressler, CTO of United Wholesale Mortgage, speaking about how AI is helping it avoid the hire-and-fire cycle the rest of the mortgage industry deals with.
"Instead of just trying to build one AI product or build a chatbot and say that we're invested in AI, I took a very different approach and said how do we weather the mortgage cycle so we never have to hire another team member, but we could do double, triple or quadruple times the business" when mortgage rates go down and business is booming, he said.
Bressler noted United Wholesale Mortgage focused its efforts on applying data extraction modeling to the underwriting function, an area of the business where there's high risk but also high reward. Not only does it not have to hire additional highly-paid underwriters, but its existing underwriters are now processing 14 loans a day. That's up from the industry standard of two or three per day. And Bressler said the number will soon be higher.
The rise of AI agents
"Today, when software shows up on our perimeter, it's a bot and it's bad. You can imagine as [AI] agents proliferate, no longer is software showing up at the perimeter bad. It actually could be a customer. It could be a partner." — Google Cloud CTO Will Grannis on how the rise of AI agents is forcing enterprises to reevaluate the way to enforce perimeter security on their corporate networks.
"One of the big big shifts happening in every large organization right now is they see the agentic future and they're trying to plan, to look at core processes and say 'how do we trust software, how do we trust agents the way we used to trust human beings?'" Grannis said during a media roundtable. "They're having to reimagine their workloads and core processes."
Model context protocol (MCP) is one of the ways the industry is trying to help solve the agent-to-data interface issue. You can learn more about that here.
Want more? Check out all our Google Cloud Next coverage here.