Snowflake CEO: Trusted Data, Not Model Benchmarks, Will Define Enterprise AI

Key Points

• Trusted, governed data will be more important than AI model benchmarks as enterprises adopt AI, Snowflake CEO Sridhar Ramaswamy said. Poorly defined or inaccurate data cannot be overcome by even the most advanced model, particularly when AI agents access sensitive business systems.
• Companies should take an iterative approach to AI modernization, focusing first on critical business functions and data rather than waiting for a complete cleanup. AI can accelerate integrations, migrations and routine analysis while employees retain responsibility for judgment and key decisions.
• Enterprise AI deployments must demonstrate measurable value and maintain spending controls, including per-user budgets. Snowflake’s goal is to use agents to reduce routine work and costs—not maximize AI usage—so organizations can improve productivity and focus on complex trade-offs.

Snowflake NYSE: SNOW CEO Sridhar Ramaswamy said enterprise AI is shifting attention away from model benchmarks and toward the quality, governance and business context of the underlying data.

Speaking at The Six Five Summit 2026, Ramaswamy argued that rapid improvements in AI-assisted software development have made code cheaper and easier to create. In his view, that change increases the strategic importance of trusted enterprise data.

“Software is getting easier and easier to create,” Ramaswamy said. “That means that if you have great data, and Snowflake has always been about getting our customers to have great trusted data governed the right way, the value that you can get from it is pretty immense.”

Data quality and governance as AI foundations

Ramaswamy said businesses can increasingly use AI to explore revenue, customer pipelines, sales effectiveness and other operating data. However, he emphasized that models cannot overcome poorly understood or inaccurate data.

“The smartest model in the world cannot make sense of truly bad data,” he said.

He pointed to the importance of knowing how enterprise metrics are defined, noting that Snowflake itself distinguishes between metered consumption and GAAP revenue because they are governed by different accounting rules. Companies need a similar understanding of their own business definitions, he said, along with controlled access and semantic context around the data.

Governance is especially important for AI agents that can access enterprise systems, according to Ramaswamy. He said a sales representative using an AI agent should be limited to information related to that representative’s own accounts rather than being able to access data on every Snowflake customer.

“Getting that right, that's not an option,” Ramaswamy said. “That is something that we absolutely have to do right.”

Modernization through an iterative approach

Ramaswamy advised companies not to wait for comprehensive, multiyear data-cleanup efforts before beginning AI initiatives. Instead, he recommended an iterative approach focused on the most important business functions and data sources.

He said AI tools can speed up system integration, pipeline development and migrations, reducing projects that once took quarters or years to weeks or months. Snowflake’s CoCo tool was among the company products he cited as helping customers accelerate those activities.

“Don't get caught in old ways of, ‘Oh, we need to clean up all our data before we can get everything done,’” Ramaswamy said.

He also encouraged companies to retain control of their data in interoperable formats, even as they use Snowflake to manage it. Enterprise applications, he said, should contribute to a shared company knowledge base rather than leave critical data isolated across separate systems or legacy environments.

• Prioritize the data and functions most central to the business.
• Bring relevant information into a governed, accessible environment.
• Attach clear definitions and meaning to metrics and datasets.
• Use AI to automate routine work while preserving human judgment for decisions and trade-offs.

Snowflake’s internal use of agentic AI

Ramaswamy described Snowflake as using its own platform internally through a centralized environment called “Snowhouse,” which he said has collected company information for roughly a decade.

That centralized view supports a sales agent that combines Salesforce data with Snowflake customer consumption information, Workday HR data and sales-enablement material, according to Ramaswamy. The goal is to provide sales teams with a more complete picture of customers and business trends.

He also described using CoCo to analyze sales outcomes. The tool can suggest analytical dimensions for evaluating won and lost use cases, such as territory, product category and AI involvement, before agents collect and analyze data in parallel. Ramaswamy said agent “swarms” can run hundreds of invocations overnight to generate reports.

Those capabilities depend on “a rock-solid foundation” of data with defined meaning and appropriate governance, he said.

Cost controls and business value

Ramaswamy said Snowflake advises customers to measure AI deployments against existing technology costs and set spending limits. He said the AI agents used by Snowflake’s sales organization cost less than the dashboarding licenses the company previously used.

“The tools that we provide for a particular function needs to cost less than what they're already using,” he said.

Snowflake supports per-user budgets, he said, allowing customers to establish monthly limits for AI tools. Ramaswamy gave an example of a customer setting an average spend ceiling of $30 per user per month for a critical application.

He said the company’s focus is not on maximizing AI token usage, but on delivering measurable value. Snowflake also allows customers to optimize their Snowflake spending through CoCo without needing to involve the company’s sales team, he added.

Looking ahead, Ramaswamy said successful agentic enterprises will use AI to handle “drudgery” such as transformation, movement of information and communication overhead. That would enable employees to spend more time on judgment, decision-making and complex trade-offs.

“The enterprises that succeed are the ones that make their organizations more effective by having AI take care of the drudgery of work,” Ramaswamy said.

About Snowflake (NYSE:SNOW)

Snowflake Inc is a cloud-native data platform company that provides a suite of services for storing, processing and analyzing large volumes of data. Its core offering, often described as the Snowflake Data Cloud, combines data warehousing, data lake and data sharing capabilities in a single managed service delivered across major public cloud providers. The platform is designed to support analytics, data engineering, data science and application workloads with a focus on scalability, concurrency and simplified administration.

Key products and capabilities include a multi-cluster, shared-data architecture that separates compute from storage; continuous data ingestion and streaming; support for structured and semi-structured data formats; tools for data governance, security and compliance; and developer frameworks for building data applications.

This instant news alert was generated by narrative science technology and financial data from MarketBeat in order to provide readers with the fastest reporting and unbiased coverage. Please send any questions or comments about this story to contact@marketbeat.com.

Continue following MarketBeat

Add MarketBeat as your preferred source on Google to see our latest stories in your feed.

Add As Preferred Source