Data Formulator 0.7: AI-powered data analytics for enterprise data

Details

Data Formulator 0.7 is Microsoft Research's open-source answer to the fragmented enterprise analytics stack. The core problem it addresses: before analysis can begin, teams must build one-off connections to each data source — databases, cloud warehouses, BI systems, object stores, local files — and redo this work every time. Data Connectors in 0.7 provide persistent, authenticated, reusable connections managed centrally, eliminating the repeated manual file uploads that slow down most team analytics workflows.\n\nOnce connected, context-aware agents assist through the full analysis pipeline: preparing and reshaping data, computing new metrics, exploring multiple analytical approaches, and generating visualizations. The system supports branching analytical workflows and maintains history, so analysts can iterate without losing prior work — a notable improvement over isolated chat interactions that start fresh each time.\n\nThe no-SQL, no-code design targets domain experts who understand the business questions but lack deep engineering skills. Analysts can iteratively refine visualizations through natural language, which the agents translate into executable operations. The interface is multimodal, supporting both text and visual feedback loops.\n\nBeing open-source positions Data Formulator as a potential standard for AI-assisted analytics infrastructure that enterprises can self-host and extend. Microsoft Research has not announced commercial packaging, but the tool is actively developed and GitHub-hosted.