'Cannot take away AI from our engineering team now': How Snowflake is embracing the new tech era

Attendees at Snowflake Summit 2025, the company's annual user conference.

Benoit Dageville, the co-founder of Snowflake, recently set off an unexpected emergency at the cloud data company - one that significantly slowed down internal productivity.

The reason? His use of Cursor, an artificial intelligence (AI) code editor that he loves and has "practically become a team" with. Dageville used it so heavily that he blew past the company's usage quota, which led to Cursor throttling access.

Everything started taking longer. Work slowed. And Dageville was up in arms. He was pacified only after Snowflake paid up for more usage.

Christian Kleinerman, the Executive Vice President of Product at Snowflake, recounted this incident to Moneycontrol while discussing how the company was using artificial intelligence internally.

"That's how critical it has become...you cannot take away AI from our engineering team now," he said, on the sidelines of the company's annual Snowflake Summit 2025.

Snowflake is also using Snowflake Intelligence internally, a newly-launched platform that enables business users to create data agents that analyse, summarise, and act on enterprise data.

Kleinerman said they have developed a range of agents to help employees in different parts of the business across the world. Among them is a customer agent which can summarise key aspects of the company's commercial relationship with a customer, including the last contract date and value, current usage, and the features in use. Other agents include those for product management, finance, and employee assistance.

"When new employees started at Snowflake, they had to open 40 tickets on ServiceNow just to get answers for easy questions. Now, the Employee Assistance agent gets you the right answer and takes you to the next step, significantly speeding up the process by weeks," said Natalie Mead, Vice President, Solutions Engineering APJ, Snowflake.

Snowflake Intelligence, which allows enterprises to ask natural language questions across structured and unstructured data through a ChatGPT-like interface, is a key piece of the company's AI strategy. At launch, it will be powered by cutting-edge large language models from OpenAI and Anthropic, running within Snowflake’s secure environment.

Kleinerman said the company is primarily focusing on the capabilities it offers to customers, rather than on which model is being used in Snowflake Intelligence.

"What matters is the value delivered, not how it's done. We're simply going to choose the model that gives us the best results," he said.

"Right now, the best models out there are from Anthropic and OpenAI, but things are changing so fast that our commitment to customers is that it shouldn't matter which model we're using. If we find something that's just as good but cheaper, we'll go with the cheaper option. If we find something materially better, we'll use that," he noted.

Kleinerman described the current capabilities of Snowflake Intelligence as an early glimpse of what’s possible. From there, it can evolve into more complex questions that go beyond quick, one- or two-minute responses.

"Imagine I'm the CFO of a company. I might say, "Go look at all of the data for the last quarter and tell me what the highlights of performance should be across everything. Look at same-store sales and product lines." You can imagine Snowflake Intelligence having a number of agents looking at different dimensions and coming back with an answer," he said.

"Today, companies have investor relations teams that do all of this. You can see much of it becoming automated," he added.

Kleinerman believes the industry hasn't yet reached a point where it's all agents talking to one another without any human involvement.

"I think anywhere there's a workflow or a business process is ripe for improvement. But I don't think there's going to be a wholesale replacement" he said.