Snowflake rolls out AI features that make data access, governance, and agent development simpler
Snowflake announced a wave of AI capabilities now generally available, with more in preview. The focus is straightforward: unify data, cut the friction of building agents, and bring natural language analysis to everyone with the right permissions.
What's GA now
- Snowflake Openflow: Automates ingestion and integration across sources and formats (including unstructured) to unify data in the lakehouse.
- Snowflake Intelligence: An AI agent that lets employees explore and explain data using natural language, not code.
- Snowflake Cortex AISQL: SQL-based tools to build scalable AI pipelines inside Dynamic Tables.
- Cortex Agents: Framework for building and governing agentic AI tools on Snowflake.
- Horizon Catalog updates: Stronger data lakehouse catalog capabilities for governance and discoverability.
- Dynamic Tables: Create AI inference pipelines directly with SQL.
- Workspaces: Shared environments to accelerate collaborative development.
- Integrations: Native ties to developer platforms such as Git and dbt Labs.
- Spark on Snowflake: Run Apache Spark code on Snowflake's engine.
New in preview or testing
- Managed Model Context Protocol (MCP) Server: Standardizes how agents interact with external data sources and models.
- Snowflake Postgres: Postgres service (including tech from the Crunchy Data acquisition) to expose live transactional data alongside analytics.
- Cortex Code: An updated assistant to interact with the Snowflake platform via natural language.
- AI Redact: Redaction of sensitive information in unstructured data within Cortex AISQL.
- Data Quality UX updates: Easier monitoring of data reliability.
Why this matters for leaders
The blocker for most AI programs isn't models-it's siloed data and fragmented governance. Snowflake's updates attack that head-on: unify data, govern it centrally, and make analysis conversational without handing out SQL privileges.
Industry analysts note two points. First, these features broaden adoption by making AI workloads easier to develop, manage, and govern-supporting data democratization. Second, they reflect Snowflake's stronger push into AI since new leadership took over, closing gaps with competitors that moved earlier.
Practical upside
- Faster time to value: Less data wrangling; more time building agents and decision support.
- Wider access-safely: Natural language analysis for business users with guardrails.
- Operational alignment: Postgres preview plus catalog updates reduce the old split between transactional and analytical data.
- Developer velocity: SQL-first pipelines, Git/dbt workflows, and Spark support reduce context switching.
Risks and questions to address
- Governance for agents: What policies control who can query what, how prompts are audited, and how results are logged?
- Data quality: Is lineage clear, and are outputs tied to certified sources in Horizon Catalog?
- PII exposure: Will AI Redact cover your unstructured data cases and regulatory needs?
- Cost control: How will you monitor compute for agent workloads and AI pipelines?
- Lock-in and portability: With AISQL and Workspaces, what's your exit or multi-platform plan?
- Skills and process: Who owns prompt standards, model selection, and change management?
Action plan for the next 90 days
- Pick 2-3 candidate use cases: finance variance explanations, supply chain exceptions, or customer health summaries.
- Pilot Snowflake Intelligence with read-only access to certified data; define approval rules and audit trails.
- Stand up a simple Cortex AISQL pipeline in Dynamic Tables and measure cycle time vs. your current approach.
- Turn on AI Redact for a sample set of emails, tickets, or documents; validate with compliance.
- Integrate Git and dbt workflows; enforce reviews for prompts, agents, and SQL changes.
- Publish a lightweight AI governance memo: access tiers, prompt logging, dataset certification, and incident handling.
- Upskill the team on agent safety, data contracts, and AI KPIs. If you need a fast start, see our AI courses by job role.
Competitive context
Databricks and the large clouds moved early on GenAI tooling. Snowflake's momentum has accelerated since 2024 under new leadership, and these releases push the platform closer to an end-to-end environment for AI apps and agents-on live, trusted data.
Where to learn more
Your membership also unlocks:
