Summary
On September 10, 2026, OpenAI launched a Data agent inside ChatGPT Work that lets employees query company data in plain language and build interactive, shareable dashboards without writing SQL. It connects to warehouses and stores including Snowflake, BigQuery, Databricks, Redshift, ClickHouse, MongoDB, and Datadog, and can pull documents from Google Drive and SharePoint.
What changed
OpenAI shipped a Data agent module in ChatGPT Work. It answers natural-language questions over structured company data, generates interactive dashboards teams can edit, share, and refresh, and can trigger follow-up actions. Native connectors include Amazon Redshift, Google BigQuery, ClickHouse, Databricks, MongoDB, Snowflake, and Datadog, plus document retrieval from Google Drive and SharePoint. ChatGPT Work is built on GPT-5.6.
Why it matters
Bringing agentic, natural-language BI into the ChatGPT Work surface puts OpenAI in direct competition with self-serve analytics tools and warehouse-native assistants, and pushes agents deeper into governed enterprise data. Respecting existing data-source permissions is the crux of whether enterprises trust it against the warehouse.
Evidence excerpt
OpenAI launched its Data agent inside ChatGPT Work, giving enterprise users the ability to pull insights, generate dashboards, and trigger follow-up actions using natural language. The Data agent connects directly to Amazon Redshift, Google BigQuery, ClickHouse, Databricks, MongoDB, Snowflake, and Datadog.