OpenAI has introduced a Data agent capability inside ChatGPT Work that aims to simplify how employees interrogate company data and create interactive dashboards using natural language prompts. The feature is designed so users need not write database queries or master analytics software to produce analytical outputs.
The Data agent can connect to approved enterprise data sources, including several major platforms: Amazon Redshift, Datadog, Google BigQuery, ClickHouse, Databricks, MongoDB, and Snowflake. In addition to direct database connections, the agent can access files and documents stored on Google Drive and SharePoint. OpenAI says the system incorporates organizational business terminology, metric definitions, and data relationships maintained in semantic layers and sources such as Databricks Genie Ontology, dbt, GitHub, Snowflake Horizon, and existing business intelligence dashboards.
Administrators at the enterprise level determine which data integrations are made available and assign permissions by role. Queries run under the connected account's current access rights, with enforcement of existing restrictions at the table, row, and column levels. That approach ensures the agent observes the same data access controls that apply to human users.
Output from the Data agent can include interactive dashboards with visualizations that teams can edit, share, and refresh. The agent is also able to build and interact with dashboards in a range of business intelligence platforms, specifically Omni, Oracle BI, Power BI, Sigma, Tableau, and ThoughtSpot. Findings generated by the tool can be distributed via Slack or email, enabling teams to circulate results through existing collaboration channels.
OpenAI reports substantial internal use: nearly all of its product organization and more than two-thirds of its go-to-market organization employ data agents in ChatGPT Work for internal analysis. The company also ran an alpha program that included participants such as NTT Data, Thermo Fisher, and ServicePiston. Those participants used the tool to analyze sales and spending, detect reporting errors, and assess business opportunities.
The Data agent is positioned as a conversational interface to company data, bridging back-end platforms and semantic metadata with an end-user experience that does not require query language skills. Enterprise controls and permissioned query execution are central elements of the implementation. Reported internal adoption and use cases from alpha participants illustrate how the feature has been applied for operational and financial analysis within organizations.