A small orchestra provided the opening prelude for Tuesday’s keynote session at Snowflake Inc.’s annual conference in San Francisco. When the music stopped and the lights went up, the conductor was revealed to be Benoit Dageville, the company’s co-founder and president of products.

Dageville’s conducting cameo provided an apt symbol for the data warehousing powerhouse as it has sought to embed generative AI across its platform and transform itself into a unified resource for intelligent data operations. Snowflake’s mission will require an ability to blend a set of disparate technologies into a harmonious whole, part of the challenge that enterprise adoption of AI has created.

“We’re beginning to think at a higher level of what the future of workflows should be,” Snowflake Chief Executive Sridhar Ramaswamy (pictured) said in a briefing for the media. “Some moments are special and the time we are going through with AI feels like one of them. There is no AI strategy without a data strategy.”

Expanding AI tools

How Snowflake intends to orchestrate its AI strategy was a central part of its message at the Snowflake Summit this week, as the company made a series of announcements designed to expand AI tools, improve processing performance and support interoperability.

The latest release of Snowflake’s core data warehouse, Standard Warehouse-Generation 2, offers a doubling of analytics performance without having to alter workloads or queries while scaling compute to handle growing data stores and complex workloads, according to the company.

EVP of Product Christian Kleinerman spoke about Snowflake’s next-gen data warehouse at Snowflake Summit

“It has faster hardware, and a lot of software optimizations,” Christian Kleinerman, Snowflake’s executive vice president of product, said during his keynote remarks on Tuesday. “We’re seeing 2.1x faster performance with analytical workloads.”

Snowflake also unveiled Cortex AISQL, the latest addition to its Cortex AI suite of offerings. Currently in public preview, Cortex AISQL allows analysts to use standard SQL commands for querying data across diverse formats including unstructured image, audio or long-form text files.

“Cortex was just an idea at our last Summit,” Ramaswamy said. “It has gone from that to become an enterprise necessity in just one year.”

The expansion of Cortex’s capabilities to encompass unstructured data highlights one of the key themes to emerge from the summit gathering. An ability to access and leverage unstructured data matters in enterprise AI, and Snowflake made a series of moves this week to address this need.

That included an announcement on Monday that the company would acquire Crunchy Data Solutions Inc., a database platform that simplifies the use of PostgreSQL without requiring management of underlying infrastructure. Crunchy Data’s vector database capabilities can make unstructured files more searchable, a feature that Snowflake believes will resonate with its customers.

“Unstructured data has gone on, almost overnight because of AI, to become dramatically more useful,” said Kleinerman in a briefing with the media.

Snowflake took another major step this week to help its customers channel unstructured data with the launch of Openflow, a managed service designed to simplify ingesting of structured and unstructured data from any source. Openflow leverages the open-tool Apache NiFi for data flow automation, with the goal of reducing the time it takes for information technology staff to manage data integration.

Multimodal capabilities for agents

Snowflake’s focus on unstructured data is driven by the firm’s plans for agentic AI. That became clearer in February when the company launched agentic capabilities for users to query combinations of structured and unstructured data using enhanced versions of its Cortex platform.

February’s unveiling included enhancements to Cortex Analyst, Search and Observability, and Snowflake expanded its agentic offerings this week. New agentic features were released across the AI Data Cloud, including Snowflake Intelligence, an agentic tool for deep research on enterprise data, and Data Science Agent, which enables users to plan and automate machine learning pipeline development.

“We’ve built very powerful multimodal capabilities for analysts to use AI,” Head of AI Products Baris Gultekin said during a breakout session at the conference on Tuesday.

Snowflake’s releases for agentic AI are following the arc of customer expectations for how the nascent technology will be deployed in the enterprise. Proprietary data estates and managerial know-how offer the prospect that large language models will enable enterprises to capture what researchers at SiliconANGLE have termed “Enterprise AGI” or artificial general intelligence.

OpenAI CEO Sam Altman discussed how agents could lead to “breathtaking” new AI models soon.

AGI is a major part of what OpenAI Chief Executive Sam Altman believes will be the “holy grail” for computing’s future. The idea is that AI can perform any economically useful function and will become increasingly more powerful, a philosophy that Altman articulated during his appearance at Snowflake’s event on Monday.

“You hear people that talk about how their job now is to assign work to a bunch of agents, that’s happening,” Altman told the Snowflake Summit audience. “We’ll start to see agents that can discover new knowledge. The models over the next year to two years are really going to be quite breathtaking.”

For Snowflake, the road to agentic AI is paved with a unified metadata model, scalability and orchestration that embraces governance, security and visibility. The company is betting that AI’s need for a robust data foundation will play to its strengths in delivering information management to the enterprise.

“We’ve made a lot of progress… to become a company that has depth in a lot of areas,” said Ramaswamy. “Our strength comes from the purity of our mission. We are centered around data. Snowflake is in a better spot because we think of AI as an accelerant to the data that people are already bringing into Snowflake anyway.”

Photos: Robert Hof/SiliconANGLE