Snowflake is jumping on the Agent train at its Build 2024 conference and launching Snowflake Intelligence. That platform should unlock data for use with autonomous AI agents.
Snowflake is launching Snowflake Intelligence at its Build 2024 virtual conference. That’s a new platform that sits atop the company’s data cloud. With Snowflake Intelligence, customers can create so-called agents. These are AI-driven processes that can take action autonomously based on the customers’ own data.
Agents in the cloud
Snowflake provides customers with a data cloud where they can keep all their data secure. In that cloud, users manage access rights to data and make it available to whom or what they want, without the data ever being copied. Data and policy rains thus remain inextricably linked.
Snowflake uses those capabilities to also give customers the ability to share certain data with third parties, or conversely integrate third-party data into their own data cloud. Think of a store powering prediction algorithms with external data about the weather. Snowflake Intelligence sits on top of that cloud platform and allows customers to build AI-based automations that work with their own data in the same secure way.
Snowflake can handle real-time and transactional data, so those are also available bia the Intelligence platform. Furthermore, Snowflake provides connections to third-party tools. Think documents in SharePoint, tools from Google Workspace, or Salesforce’s solutions including Slac. Snowflake Intelligence further supports APIs, allowing users to connect an AI agent to other external tools as well.
Focus on unlocking information
The new platform will allow users to ask questions about all their data, after which agents go to work looking for answers through all the different sources. Agents are more than co-pilots, implying that they can also take action on the data. The focus of agents at this point does seem to be on analysis, summaries and text generation. This is a narrower view of the concept than that of Salesforce, for example, where agents can be given real concrete tasks based on the data, similar to that of a human.
read also
No Clippy, but Sophie: AI according to Salesforce passes the (Microsoft) Copilot
Snowflake Intelligence is built on Snowflake Cortex AI, which in turn is the system that securely binds data to AI tools. Central to the whole story remains the management layer, which ensures that data is never copied, and access and security rules always remain in place. So it’s not like sensitive data is suddenly accessible through a workaround because you’ve tied it to an AI solution.
The right direction
Snowflake Intelligence is a relevant but also necessary solution for Snowflake. Data and AI are two chapters in the same story. Snowflake offers an excellent data foundation that makes much possible, but that approach also has limitations. Those in the Snowflake ecosystem are limited to the capabilities of that ecosystem. In other words, Snowflake cannot be left behind in terms of AI functionality, because then the data cloud suddenly becomes a data prison.
With Snowflake Intelligence, the company is unlocking data in the data cloud for customers who want simple but powerful AI functionality. That aligns with the Salesforce vision we heard earlier this year, although Snowflake’s definition of agents does seem narrower than Salesforce’s at the moment. However Snowflake is rapidly carpeting out the right solutions. After all, a lot of AI-related functionality was already possible with Cortex AI. Snowflake Intelligence will soon be available as a private preview.