Snowflake Summit: The Path to AI Success With Snowflake

“Start small, prove trust, then scale.”

That’s the advice CDW’s Jay Brophy had for leaders looking to achieve artificial intelligence success with Snowflake’s AI Data Cloud.

In a presentation at Snowflake Summit 26 in San Francisco, Brophy noted that organizations should view AI not a product to be purchased and used, but as an operating capability that changes how users access data, who can use that data and how quickly decisions get made. Some observers have expressed concern that AI will lead to job losses, but Brophy stated that the true potential of AI lies not in replacing workers but in enhancing their abilities. 

“CoWork isn’t replacing your analyst,” Brophy said, referring to the name Snowflake announced for the rebranding of Snowflake Intelligence, which is a personal work agent for knowledge workers, not just a chatbot. “It’s shortening the time it takes to get answers.”

To maximize the AI capabilities of Snowflake’s AI Data Cloud, organizations need to overcome common hurdles, Brophy noted. AI and natural language analytics projects fail far more often because of data governance issues and semantic ambiguity than because of problems with the AI technology itself.

Data readiness represents a significant barrier to AI adoption. Many organizations rush to deploy AI interfaces before agreeing on metric definitions. To overcome these hurdles, organizations should set clear rules for data governance and establish a semantic layer. This layer defines key performance indicators as well as other metrics and terms such as “revenue” or “customer,” so all users and AI tools are working with well-governed data. 

“The semantic layer also provides guardrails, defining what the model can infer and what it can’t infer,” Brophy said.

Trust is a make-or-break factor for enterprise AI success, Brophy noted. Users may tolerate slower access to data, but they will not tolerate inaccurate answers. In fact, an AI tool that provides wrong answers or other bad responses can permanently damage confidence in the system and the data platform owners.

Users need to trust their AI tools and the data supporting them. To achieve this, organizations should conduct accuracy testing, implement robust data governance, define data owners and establish data lineage. Once users know how AI is reaching its conclusions and validating the accuracy of those results, they can begin to trust it. 

“Trust isn’t a feeling,” Brophy said. “It’s earned through traceability.”

Brophy outlined four stages that organizations should follow to deploy AI successfully. They include:

• Semantic readiness. At this stage, organizations must define the metrics their AI systems will use and align stakeholders on these metrics.
• Governance prerequisites. This stage establishes data ownership and lineage while laying out access controls. 
• Activation. Many organizations make the mistake of starting with this stage, Brophy stated, but to be successful, they must complete the first two stages before queries and analytics begin.
• Adoption. This stage includes building trust as well as onboarding users. The Snowflake AI Data Cloud is particularly valuable here, Brophy noted, as it helps users take advantage of robust AI capabilities such as Snowflake’s CoWork interface. “The bottleneck of every AI deployment is skilled workers,” he said. “Snowflake’s Cortex Code helps close that gap. It’s a practitioner accelerator.”

Cortex Code, which Snowflake rebranded as CoCo during the summit, is a powerful tool to support AI adoption. CDW’s status as a winner of the Snowflake CoCo Champion Badge, demonstrates its expertise in building AI-powered applications, automating complex workflows and accelerating end-to-end development.

Ultimately, organizations will succeed with AI when they establish governance, deliver clear metric definitions and gain users’ trust. All of this should happen before exposing any data to a natural language interface. 

As they plan their AI implementations, organizational leaders must understand that AI technology is the final layer, not the foundation for success. When organizations overcome common hurdles by using the Snowflake AI Data Cloud, the results accelerate, improving innovation and time to value.

“When a user starts their day asking CoWork what happened yesterday, decisions and actions quickly follow,” Brophy said.