Google Cloud CEO Thomas Kurian said AI agent interoperability is critical to link process and workflows, outlined the sovereign AI strategy and said the company is building an agentic AI ecosystem.

Speaking at an ask me anything session with analysts, Kurian said the following at the end of day 1 at Google Cloud Next.

Scaling the organization while looking around corners. "Our strength as an organization has already come from one understanding deeply what customer problems we're trying to solve," said Kurian. "We are very careful, given the number of problems out there, to focus on a very specific set of problems. We focus on a few important ones."

Building the go-to-market ground game. Historically, Google has been a company that has had great technology, but difficult to buy and deploy. That vibe has changed as Google Cloud has built an ecosystem. Kurian said Google Cloud can appeal to multiple CxOs. Kurian said: "If you look at AI, much of the buy decision is not in IT. It's outside it. How you engage not just in IT, but outside, and how you can build a solution portfolio that our team will sort of take market is constantly evolving."

Process flow and accuracy. Kurian was asked about the success of agents given hallucinations and how enterprises may be reluctant to hand off processes to AI. "We've taught people to decompose. Take a process and break it into a set of sub steps. We need people to go through each step and not design one agent. You need to have that ability to view each step and prepare the accuracy of that step," said Kurian. "Enterprises want predictability."

Starting with agentic AI. Kurian said companies just starting out need to start with the business problem to solve and measure the result. Data quality feeding that AI agent is also critical. There are also industry considerations. Workflows and security are also issues. Kurian said: "There's a lot of specific guidance whenever we talk to customers. The best proof point is always look at others in the industry and what they may have done so that you can actually face the timescale you think you need to change the organizational change management."

Why Kurian seems happier? First, Chappell Roan called him back (even if she isn't performing at Next). That call was noted during the keynote. "I've always been proud of our team. Many people thought we would not succeed in it. Many people thought that we would never be successful outside of the consumer domain. I'm proud of the team and the resilience we've shown."

Solving big showcase problems. Kurian said "The Wizard of Oz" project is an example of how Google Cloud likes to participate in big projects, but let the customers tell the story. "I don't think people realize Dorothy shoes only show up in eight frames so tuning a model takes a lot of work to improve the model. We chose to work with that because if you can do this here, you can do pretty much anything else," said Kurian. Google Cloud is also working on similar tough projects with hedge funds, healthcare, life sciences and manufacturing. "We like to change the way those industries work," he said.

DeepSeek. Kurian: "The industry tends to rotate every few months because a bunch of articles that were written that may not be accurate. They've done some really good things. If you look at the cost of our models, I think people misinterpret why we didn't say anything. We're very confident that the actual training cost for a model from us is comparative to anybody out there."

"We always say there's a lot of debate on these topics. And if you actually sit down with the DeepMind guys who are working on it, I think you see the reality of some of the numbers. That's why, when we don't say something, it's not that we are surprised and it's obvious what they did. Credit to them for doing it."

 

Business Research Themes

Data to Decisions

Roles

Chief Information Officer

Tags

@googlecloud, #AI #artificialintelligence #machinelearning #ML #generativeAI