Snowflake VP of AI Baris Gultekin on Bringing AI to Data, Agent Design, Text-2-SQL, RAG & More - Cognitive Revolution Recap
Podcast: Cognitive Revolution
Published: 2026-01-14
Duration: 1 hr 39 min
Guests: Baris Gultekin
Summary
Baris Gultekin discusses the transformative role of 'bringing AI to the data' in enterprise settings, emphasizing security, governance, and the importance of high-quality retrieval. He covers Snowflake's AI strategies, including text-to-SQL advancements, agent design, and the company's approach to open standards.
What Happened
Baris Gultekin, Vice President of AI at Snowflake, explains the company's innovative approach to integrating AI with enterprise data. By bringing AI to the data rather than sending data to model providers, Snowflake enhances security and governance, a critical consideration for enterprises handling sensitive information.
Gultekin highlights the significant advancements in text-to-SQL technology, which are making natural language data analysis more reliable for business users. He notes that Snowflake's semantic modeling improvements and the integration of Neva's AI-powered search capabilities are key to this process.
The episode delves into Snowflake's approach to Retrieval-Augmented Generation (RAG) systems, emphasizing the critical role of embedding models and chunking strategies. These systems enable businesses to efficiently handle large volumes of unstructured data, which comprises 80-90% of enterprise information.
Snowflake Intelligence, the company's agent platform, is gaining traction as a tool for business users to leverage AI for data-driven decision-making. Gultekin underscores the importance of model choice and cost tradeoffs, as well as the need for strict governance and security measures.
The conversation touches on the competitive dynamics in the AI industry, with Gultekin noting that as model quality becomes more uniform, the application layer will be crucial for differentiation. Snowflake's partnerships with companies like Anthropic and Google for Gemini models are part of its strategy to offer diverse AI solutions within a secure boundary.
Gultekin discusses the potential of narrow AI models, which may offer more control and safety compared to broader general AI systems. He also emphasizes the democratization of data access and insights, a shift that AI is accelerating across enterprises.
Finally, the episode explores the challenges of getting enterprise data AI-ready, including breaking down silos and making data more accessible. Snowflake's focus on ease of use in deploying AI at scale remains a core design principle, aiming to help businesses quickly capitalize on AI's capabilities.
Key Insights
- Snowflake enhances enterprise data security by integrating AI directly with data, eliminating the need to send sensitive information to external model providers.
- Advancements in text-to-SQL technology, supported by Snowflake's semantic modeling and Neva's AI-powered search, are making natural language data analysis more reliable for business users.
- Retrieval-Augmented Generation (RAG) systems in Snowflake utilize embedding models and chunking strategies to efficiently manage the 80-90% of enterprise data that is unstructured.
- Snowflake's partnerships with companies like Anthropic and Google for Gemini models aim to provide a diverse range of AI solutions while maintaining strict security boundaries.