Snowflake World Tour 2025 – “Enterprise AI”

29 Aug 2025 Snowflake World Tour 2025 – “Enterprise AI”

I once again had the opportunity to head to Snowflake’s world tour in Sydney, and thought I would take the time to summarise some of the things I took away from the conference. If you’re interested in what I thought last year, you can read that here

The Era of Enterprise AI and Snowflake

Following on from the announcements of AI tools and features last year, this year the title was “Era of Enterprise AI,” with a focus on empowering more people to access AI tools and data. They released an impressive 125 AI-related features in the last 12 months, including Snowflake Intelligence and enhancements to Cortex AI. Their goal is to help enterprises move from ”specialised, compartmentalised, data usage to a more democratised, and augmented approach”. Or to put it more simply, getting non-technical people generating meaningful insights with the help of AI. The hype and promise of AI was tempered by the repeated mantra of “There is no AI strategy without a Data strategy”, and that’s certainly a point I can agree with.

Cortex AI: The Core of Snowflake’s AI Strategy

Cortex AI is positioned as a powerful tool with 90% accuracy (self-reported). The keynote featured the experience of Commbank iQ’s CEO using it as an AI-powered assistant. There have been several enhancements to Cortex AI this year including:

• Cortex AI SQL: A new query language that enables users to work with AI functionalities directly within SQL. It can also provide references to verify answers.
• Cortex Agents: These agents can run both inside and outside of Snowflake and support custom tools and memory. They can be created through a platform and are designed to help answer key business questions, as well as undertaking workflows.
• Snowflake Intelligence: A new chat based UI that allows for natural language queries, powered by Snowflake Cortex Agents, Analyst and Search. If you’ve used ChatGPT, Gemini or Claude you’ll be right at home here.

A key feature for me of Cortex AI is that Snowflake runs models where the data resides, with built-in governance and the guardrails that already exist in your environment. The key benefit of this is that the Agents and tools have the same access to the underlying data as the user using them, leveraging the strong foundation of row and column security, and masking policies that organisations have already built up.

Snowflake’s Performance and Migration Tools

The conference highlighted significant performance improvements and migration capabilities:

• Gen 2 Warehouses: These warehouses are noted as being 2.1 times faster than previous versions.
• Snowpark Connect: A feature for Spark that supports existing Spark workloads and includes an AI-powered migration toolkit.
• Snowconvert: This is a key tool for accelerating data warehouse migrations. It boasts significant savings and time reductions, with one example citing an 88% reduction in conversion time and another mentioning an 84% saving after migration. Snowconvert supports migrations from various platforms like Redshift, BQ, Oracle, Teradata, and SQL Server. It can automatically create test cases and provides a VS Code extension.

New Capabilities: Dynamic Tables, OpenFlow

Several other new features and enhancements were discussed that would drive real quality of life improvements for those moving to/managing Snowflake:

• OpenFlow: is a managed ingestion feature for both structured and unstructured data, built on top of Apache Nifi. You can leverage it not just as an ingestion tool but also orchestrate data moving out of and around Snowflake. They have thought through this, and offer a couple options already: Snowflake managed and hosted, and self-hosted in AWS. There is a planned On-prem option in the works. The on-prem option piqued my interest as a great way to get data from legacy systems into Snowflake.
• Dynamic Tables: A long standing and cool feature of Snowflake got some nice usability and quality-of-life improvements, including immutability zones, insert-only inputs, backfill, and storage lifecycle policies.

Wrapping Up

The conference was really well attended with an estimated 3000 attendees, and a pretty packed partner expo. It was great to hear about real use cases for the AI features being highlighted from people who had actually lived with them. It certainly left me with more than a few use cases that I could apply both internally within Shine and to a few of our Snowflake customers.