How Snowflake Achieved 19.5% Higher Feature Adoption with AI-Powered Account Intelligence

From strengthening supply chain operations to improving healthcare staffing outcomes and saving 54% of costs, we’ve seen how Snowflake AI tools are transforming outcomes for our customers around the world. 

That same transformative power is at work within Snowflake, too. For example, what if a marketer could predict which of their thousands of customers would benefit from a specific product feature before the customer even knew they needed it? Discover how in one quarter (February-April 2025) Snowflake's marketing team saw a 19.5% increase in feature adoption among targeted accounts and a 45% higher meeting-to-conversion rate for our sales team, using LLMs and graph neural networks (GNN) with Kumo.ai. 

AI-driven insights for customer marketing 

Every marketer knows the challenge of prioritizing product messaging at the right time to the right audience. At Snowflake, the lifecycle marketing team turned to AI to see how they could use their own data to help with campaign prioritization, executing efficiently and at scale. 

A truly collaborative effort, the product marketing and management teams, go-to-market and experimentation team, and greater marketing organization worked together to identify product signals and triggers that could create a successful model to predict a customer’s next Snowflake use case. 

The challenge: Scaling personalized B2B marketing

As with most businesses, Snowflake’s success relies not only on growing the number of customers but also helping them discover new products and features that support their goals. Snowflake’s traditional lead scoring told the team who might buy, but not what they should buy next — a costly blind spot in our expansion strategy. Traditional tools for account intelligence looked at contact-level details and lead scoring, primarily focused on net-new business, not new product adoption. 

“Traditional scoring gives you a yes or no: Will this lead convert or not? We needed to answer which of our use cases should this account adopt next, and in which order,” says Daniel Chow, Senior Data Scientist at Snowflake and member of the Marketing AI Council. “We had all the signals — product usage, event attendance, feature exploration — but manually analyzing patterns for thousands of accounts across dozens of potential use cases wasn’t just complex, it was impossible without AI.”

What the team was looking to build for customer intelligence was more complex, pulling in multiple data sources, data types and marketing intelligence tools. Faced with many discrete outcomes to predict and complex data to ingest, the Snowflake team tapped into the Snowflake Native App network and found a solution with Kumo AI, an AI model company focused on relational data. 

Solution: From traditional ML to graph neural networks

Working with a Snowflake Native App was beneficial for many reasons. The seamless, instant integration meant the team didn’t have to worry about data movement, governance or data security, and it made the often lengthy back-and-forth with procurement much smoother. 

“The Snowflake Native App integration meant we could leverage our existing Snowflake infrastructure without any data movement required,” says Syed Zaidi, Data Scientist Marketing Intelligence at Snowflake. “We pointed it at our tables and within 48 hours had predictions that outperformed two to three months of traditional model development.” 

Data was pulled from multiple domains with hundreds of millions of product telemetry records and tens of millions of sales activity records. All of that data, both structured and unstructured, was ingested into the GNN for account identification. Being able to use such a comprehensive amount of data led to a significant performance boost compared to previous efforts. 

Kumo was a great choice for this task, as well, because of its use of graph neural networks. GNNs excel at finding patterns in connected data; instead of analyzing customer behaviors in isolation, like traditional ML does, GNNs look at the influences and shared characteristics, learning through relational data.  

“What impressed us was how the GNN handled the complexity of our data. We fed it credit consumption patterns, tool usage, campaign responses, sales notes, feedback and more. Over 100 million records that would have been impossible to feature engineer manually,” says Matt Loskamp, Senior Manager of Data Science at Snowflake. “The model discovered patterns we never would have found, ultimately achieving 20% better account prioritization than our traditional approaches.”

With the native integration into Snowflake, predictions can be refreshed daily instead of quarterly. 

From pilot to global success

Armed with this data and hypertargeted, AI-generated messaging, the lifecycle marketing team launched customer campaigns to run over the course of a quarter. Throughout the three months, email drip campaigns, product-specific events, SDR outreach and more supplemented the retargeting efforts.The resulting campaign saw a lift in engagement with a 19.5% increase in feature adoption among targeted accounts and a 45% higher meeting-to-conversion rate for our sales team. The success of the pilot campaign has now scaled to global execution across multiple workloads, changing how we approach customer marketing at Snowflake. 

Ready to transform your own B2B marketing with AI? Join our webinar to dive deeper into the technical implementation, see the complete campaign journey and learn how you can achieve similar results with your own customer data.