2 min Analytics

dbt Labs launches Fusion engine and AI agents for data teams

dbt Labs launches Fusion engine and AI agents for data teams

The new tools promise cost savings and accelerated development for data platforms.

The company announced this during its Coalesce 2025 conference. dbt Labs is bringing four different AI agents to its platform, each designed for a specific part of the data development process. The Developer agent explains logic, flags duplication, and helps refactor code in VS Code or dbt Studio.

The Discovery agent helps find the right datasets and definitions, while the Observability agent monitors jobs and identifies possible causes of problems. Finally, the Analyst agent answers questions about models, jobs, and metrics within dbt Insights.

Cost optimization through smarter orchestration

The dbt Fusion engine, now in preview for BigQuery, Databricks, Snowflake, and Amazon Redshift, promises significant cost savings for organizations. State-aware orchestration reduces compute costs by approximately 10 percent by only executing models that have actually changed.

Organizations can achieve further optimizations by specifying specific data freshness requirements. The engine then determines the most efficient execution path. In early tests, some organizations have achieved more than 50 percent cost savings.

The Fusion engine extends support to Apache Iceberg tables in Snowflake and Databricks. This facilitates the adoption of open table formats and cross-platform portability for data teams.

The dbt VS Code Extension, now in preview, allows developers to run Fusion locally for better integration with production environments. dbt Insights is now generally available, bringing definitions, lineage, costs, performance, and reliability together in one platform.

MetricFlow becomes fully open source

dbt Labs announces that MetricFlow is now available under an Apache 2.0 license. This step underscores its commitment to the Open Semantic Interchange (OSI) initiative for standardizing metrics and semantics across different tools.

Tristan Handy, founder and CEO of dbt Labs, emphasizes that open standards and AI are driving the next era of analytics. “Fusion delivers robust context, tools and error-correction mechanisms for both humans and agents. It is the enabler of next generation, AI-powered data infrastructure.”