OpenAI has launched the Data agent in ChatGPT Work, a plugin designed to let employees query corporate systems, surface supporting evidence and produce shareable dashboards using ordinary language. The Data agent in ChatGPT Work aims to lower the barrier to analytics by removing the need to write SQL or learn new BI interfaces, and by interpreting an organization’s existing business definitions and semantic layers.
The agent connects to a broad set of approved enterprise data sources. OpenAI lists integrations for Amazon Redshift, Google BigQuery, Snowflake, Databricks, ClickHouse, MongoDB and Redis, and the tool can also draw on files stored in Google Drive and SharePoint. Administrators control which connections are available to their organization and can govern role-based access; queries executed through the agent respect account-level permissions including table, row and column restrictions.
OpenAI says the Data agent leverages an organization’s semantic layers — examples cited include dbt, Databricks Genie Ontology, GitHub and Snowflake Horizon — to interpret metrics and relationships in line with existing business definitions. That alignment is intended to ensure that answers reflect the same metric definitions and logic that teams already use in internal reporting, reducing the need to re-validate outputs against established sources.
Users interact with the agent via natural-language prompts and follow-up questions within the same conversation. The tool surfaces the evidence behind its findings, allows users to refine analyses, and can generate interactive dashboards with built-in visualizations. Those visual outputs can be exported or published into BI tools already in use by teams, and the agent can adapt visuals to an organization’s brand guidelines where required.
OpenAI lists an array of BI integrations, including Tableau, Microsoft Power BI, ThoughtSpot, Sigma, Omni and Oracle BI. The Data agent can either publish dashboards directly into those platforms or build views within them, helping teams move from insight to action without manually reproducing visuals. In addition to producing dashboards, the agent can recommend next steps, identify relevant stakeholders, share results via Slack or email, and perform approved actions through connected tools.
The company positioned the public release alongside statements from platform partners and early customers. Platform partners named include AWS, ClickHouse, Databricks, Snowflake, MongoDB, Redis and G2. OpenAI also described an Alpha program involving organizations that used the agent for tasks such as sales and spending analysis, detecting errors in reports, and prioritizing opportunities. Customer organizations cited by OpenAI include NTT DATA, Thermo Fisher Scientific, ServiceTitan, Zipline, Empower, Doeren Mayhew, CookUnity, Turing, micro1 and Unit8. According to OpenAI, those teams used the agent to build dashboards, validate findings against internal reports and reduce manual reporting work.
OpenAI says the tool draws on its own internal practices: the company’s product and go-to-market teams already use data agents in ChatGPT Work, and OpenAI’s data team developed shared business definitions, access controls and safeguards for sensitive data to support internal use. The company also links to additional implementation materials, such as a post from its data team and a webinar for administrators and users who want deployment guidance.
Administrators can enable the plugin from the Plugins directory in ChatGPT Work under the listing named Data. Organization admins can configure specific data-source plugins (for example, Databricks and Snowflake) in Workspace settings > Plugins, and they can manage which users or roles have access to each connection. If the plugin is not yet installed, OpenAI instructs users to add it from the Plugins directory and complete any required account-connection steps before beginning a conversation with @Data.
To illustrate common workflows, OpenAI provides sample prompts that show how the agent can diagnose metric changes, design KPI frameworks for product areas and prepare leadership-ready metric readouts. Typical use cases highlighted include adoption and retention analysis, operational performance reviews, product growth assessments and financial analysis.
By combining direct data connections, existing semantic layers and conversational interaction, the Data agent in ChatGPT Work is positioned as a way to make analytic workflows accessible to non-technical employees while preserving governance and permission models. The ability to generate, validate and publish dashboards into existing BI platforms is intended to accelerate insight-to-action cycles and reduce the manual effort involved in routine reporting.
OpenAI published the Data agent announcement on September 10, 2026, and made implementation resources available for organizations evaluating deployment. The release signals a step toward embedding natural-language analytics into enterprise workflows while emphasizing controls and alignment with established business definitions.
Source: Read the original source
Leave a Reply