Applied AI

When Snowflake CEO Sridhar Ramaswamy and CFO Brian Robins face Wall Street analysts on quarterly earnings calls nowadays, they’re armed with material from an internally developed AI agent. The agent tells the two executives the questions they’re expecting the analysts to ask and provides suggested answers.

The agent prepares that material in minutes, automating a process that used to take Snowflake staffers several weeks, according to Anahita Tafvizi, the company’s chief data and AI officer. And it’s one of the ways Snowflake employees are using AI to make them more productive. And by themselves getting better at using AI, Snowflake is getting better at selling AI to its customers, according to a current Snowflake salesperson. 

Tafvizi's team works with multiple Snowflake business units, such as sales and marketing, finance and human resources, to develop AI agents that automate manual work. They do this using Snowflake’s two main AI products: an AI coding agent called Cortex Code and an agent for querying data in Snowflake and third-party applications, called Snowflake CoWork. 

Aside from the agent coming up with analyst questions, another agent developed for Robins tracks Snowflake customers that are deviating from their projected levels of spending—both positively and negatively. It analyzes the accounts to figure out why, and drafts emails with sets of questions for the salespeople handling the accounts. “All Brian has to do is review the emails and send them,” she said. 

Snowflake has also developed personalized agents for Chief People Officer Arnnon Geshuri and other senior executives, Tafvizi added. 

Ramaswamy in all-hands meetings frequently emphasizes to employees the importance of using its AI products, said the salesperson. His message is that employees need to be using AI daily in the course of their work or risk falling behind, the salesperson said. 

To encourage AI use, Snowflake provides salespeople with “unlimited” access to its AI products powered by any model the company offers to customers, including high-end ones from providers like Anthropic and OpenAI through Microsoft Azure, the salesperson said. A Snowflake spokesperson didn’t have a comment. 

Many Snowflake salespeople are building custom applications to help them hit sales targets and find ways to help customers save money and potentially get them to move more of their data into Snowflake, said the salesperson. 

There are signs the effort is working. The number of customer accounts using Snowflake’s AI passed 13,600 last quarter, up from more than 9,100 accounts the previous quarter, Snowflake revealed in earnings press releases. 

Snowflake often highlights its internal use of AI in presentations to customers, said the person. In cases where customers want to know about how to manage AI costs, Snowflake salespeople will show them how its own teams navigated the costs of using AI models. Or if a customer asks how to develop agents that handle finance tasks like processing invoices, salespeople will highlight how Snowflake’s own finance team carried out this work, the salesperson said.

“We are using AI in every corner of the business,” said the salesperson.

A message from Google Cloud Navigating the next era of AI infrastructure. Building for agentic AI requires a new approach to hardware and software. Tune in to Amin Vahdat, SVP and Chief Technologist, AI and Infrastructure at Google, as he walks through key shifts in AI infrastructure, sharing practical insights on how system-level co-design can help you scale your workloads efficiently. Watch the keynote.