Snowflake Tries to End the AI Data-Access Wars
Snowflake is reversing an incendiary trend of software incumbents such as Salesforce and Atlassian walling off data in their applications from artificial intelligence startups that threaten them.
On Tuesday, Snowflake said a consortium of more than a dozen other enterprise app and database providers—including, ironically, Salesforce—have committed to operating business intelligence and other analytics apps so that chatbots or other conversational AI can pull data from all of them at once.
For instance, a salesperson that’s about to meet with a customer might want to know about that customer’s spending habits, whether they might be a good candidate to try a new product, and whether they have unresolved customer-support problems.
Right now, such information might be held in several different databases or applications at the salesperson’s company, such as Salesforce’s Tableau or data “catalog” firm Alation, said Josh Klahr, a product management leader at Snowflake, which is known for selling databases.
And the data stored in those services comes in many forms, making it hard for conversational AI to know what data to grab. The consortium, which also includes business intelligence provider ThoughtSpot, aims to ensure the data—including everything from numbers to nouns and verbs—are stored in a standardized format the AI can understand.
AI-powered search is already one of the hottest applications of the technology in the enterprise, and one such provider, Glean, was an early breakout success.
But Salesforce’s Slack as well as Figma, Notion, Atlassian and others have been making it harder for Glean and enterprise software customers to execute cross-app searches with AI, as we have reported in recent months. Even OpenAI, which also wants to enable ChatGPT customers to access data across numerous productivity apps, has been caught in the crossfire.
Klahr said enterprise software customers expect more flexibility from their vendors, and new standards such as Model Context Protocol will make it even easier for AI tools to find and use data across all types of apps, no matter where the data are stored. (MCP eliminates the grunt work that developers of agents need to do for AI to get information from apps or take actions involving them, such as making a purchase or spinning up a marketing campaign.)
“Our customers are very interested in openness and interoperability—bets that don’t lock them into” a particular software vendor forever, Klahr said. One Snowflake customer, asset management giant BlackRock, was listed in the initial members of the consortium.
Of course, the Snowflake-led consortium represents only a small fraction of the BI and analytics market, so it’s largely a symbolic gesture at the moment. We asked spokespeople from Google and Microsoft, which controls the most popular BI app, what they thought of Snowflake’s effort but they didn’t provide an immediate comment.
Why Large Companies Are Struggling With AI
Hemant Taneja, CEO of General Catalyst, believes many large companies aren’t getting a return on their AI investments because they haven’t developed a comprehensive strategy for adopting it.
Some of the challenges are technical ones. To effectively use AI, companies need to organize and prepare their data and then train AI models on it, Taneja said in an interview Wednesday on TITV, The Information’s new streaming show.
Adopting AI also requires firms to rethink existing management structures to account for situations where AI will manage human employees—or humans will manage AI agents. Senior leaders also need to show they’re committed to all of these changes, he said.
In a recent study, the Massachusetts Institute of Technology claimed that 95% of corporate efforts to use artificial intelligence failed to boost the bottom line for the surveyed companies. According to Taneja, such disappointments happen because companies aren’t following all of these necessary steps.
“They can see what’s possible, but then they get stuck,” Taneja said in the TITV interview. “You have to have all four of these ingredients to really [drive adoption and get value].”