OpenAI has launched a Data agent in ChatGPT Work that lets any employee connect to company data sources and get interactive dashboard answers in plain language — no SQL, no BI training, and no data analyst required. The agent is available now as part of OpenAI’s enterprise product expansion.
What the ChatGPT Data Agent Does
The Data agent is available as a plugin in the ChatGPT Work Plugins directory. Employees type a plain-language question — “Which campaigns drove the most pipeline last quarter?” — and the agent queries the connected data source, applies the organisation’s existing semantic layer and permission rules, and returns an interactive result.
Supported data sources include Amazon Redshift, Datadog, Google BigQuery, ClickHouse, Databricks, MongoDB, Snowflake, Google Drive, and SharePoint. The agent reads business logic from semantic layer tools — Databricks Genie Ontology, dbt, GitHub, and Snowflake Horizon — so results use the company’s own definitions of metrics like “revenue” or “churn,” not the agent’s defaults.
Administrators control which data sources are available and who can access them. The agent enforces the connected account’s existing row-level, column-level, and table-level permissions. An employee cannot access data they could not already access through their database credentials.
Results can be shared via Slack or pushed directly to BI platforms including Omni, Oracle BI, Power BI, Sigma, Tableau, and ThoughtSpot. The agent does not replace those platforms; it adds a natural-language query layer in front of them.
How Businesses Are Using It
Three companies reported active deployments at launch.
NTT DATA reduced the technical barrier to dashboard creation. According to Yuji Shono, Head of Global AI Office at NTT DATA Group: “Many non-engineers, particularly in sales and corporate functions, have been able to build and update their own dashboards using plain language.” Shono cited licensing costs and technical expertise as previous blockers to expanding dashboard access across the organisation.
ServiceTitan used the agent to discover a performance difference inside its product. A dashboard built through the Data agent revealed that users of its Atlas AI sidekick launched marketing campaigns 3 times faster than non-users — a finding that validated the product investment with a specific metric.
micro1 rebuilt its performance-tracking dashboards in 30 minutes using the agent, and identified errors in the original dashboards during the process.
OpenAI reports internal adoption at scale: nearly all of its product team and more than two-thirds of its go-to-market organisation use data agents.
What This Means for Business Software Buyers
OpenAI launched the agent on September 10, 2026. The Data agent is available only on ChatGPT Work — the Business, Enterprise, and Education plans. Teams already paying for ChatGPT Work get access through the Plugins directory without a separate analytics license.
For business buyers, the relevant competitive question is not whether the agent replaces a dedicated BI tool. It is whether any CRM, project management, or workflow platform that still requires a trained analyst to extract insights remains easy to justify when a conversational query layer now comes bundled with the productivity suite their teams already use.
The Data agent also raises the floor for what counts as a useful AI agent for business: a tool that answers questions in chat is now table stakes, while a tool that queries live data, enforces permissions, and pushes results into existing dashboards is the new baseline for enterprise-grade AI. For a broader view of where AI fits across business operations, see our guide to AI tools for business.
Our Take
OpenAI’s Data agent marks the clearest shift yet from AI as a question-answering tool to AI as an operational analytics workflow. The NTT DATA quote is direct: the previous blockers were licensing cost and technical expertise, not data quality or model capability. The Data agent removes both. Every business software vendor whose platform still requires a dedicated analyst to surface insights is now competing against a capability bundled into ChatGPT Work at no incremental cost. That is not a feature gap — it is a pricing and positioning problem that will compound with each enterprise renewal cycle.
