OpenAI ships Data agent in ChatGPT Work for company data and dashboards

10 Sep 2026: OpenAI published “Now everyone can put data to work” — a Data agent in ChatGPT Work that connects approved company sources, investigates changes, and builds shareable interactive dashboards via natural language. Company post is the primary. Alpha/customer usage figures stay OpenAI’s.

10 Sep 2026: OpenAI published “Now everyone can put data to work,” introducing a Data agent in ChatGPT Work so people can turn company data into answers, interactive dashboards, and approved actions by asking in natural language. That dated openai.com post is the filing event.

Connectors named on the post: approved sources including Amazon Redshift, Datadog, Google BigQuery, ClickHouse, Databricks, MongoDB, Snowflake, and more; also files and documents from Google Drive and SharePoint. Semantic and context sources cited include Databricks Genie Ontology, dbt, GitHub, Snowflake Horizon, and BI dashboards.

Admins choose which connections and roles can use the agent. Queries, OpenAI says, enforce the connected account’s existing table, row, and column permissions. That is OpenAI’s stated control model — not a desk audit of customer tenancies.

Analysis flow on the post: ask follow-ups, review evidence behind findings, and turn analysis into interactive dashboards that teams can edit, share, and refresh. The agent can also build and interact with dashboards in Omni, Oracle BI, Power BI, Sigma, Tableau, and ThoughtSpot. OpenAI says it can recommend next steps, share via Slack or email, and carry out approved actions through connected tools.

Getting started on the post: the Data plugin in the ChatGPT Work Plugins directory; admins can make it available or install it via Workspace settings > Plugins; start a conversation with @Data. Example prompts on the page stay illustrative — not measured results.

OpenAI’s own usage claim on the post: nearly all of its product team and over two-thirds of its GTM organization use data agents in ChatGPT Work. Alpha customers named: NTT Data, Thermo Fisher, ServicePiston, and “other organizations.” Treat those as OpenAI’s Alpha-program claims, not independent audits. This filing does not invent pricing, GA versus Enterprise-only availability, or a claim that the agent replaces BI teams — those do not appear as hard claims on this primary. Distinct from the already-filed shorter openai-chatgpt-work-data-agent launch note, and from openai-documents-agents-api, openai-published-gpt-6-astra, cognition-devin-gpt6-astra-testing, openai-chatgpt-financial-services, openai-habitat-storage-one-billion-users, and google-tunix-autofinetune.

A dated OpenAI primary puts a governed Data agent inside ChatGPT Work — desk keeps connectors, admin permissions, and Alpha usage as company claims, not proof that every org can skip analysts.