From a pending report to a direct question
OpenAI has launched Data agent inside ChatGPT Work. A user can ask why sales slowed or where spending rose, and the agent searches approved sources, investigates the change and builds an interactive dashboard without requiring SQL.
The official announcement lists Redshift, BigQuery, ClickHouse, Databricks, MongoDB, Snowflake and Datadog, plus files in Google Drive and SharePoint. It cannot rescue undefined metrics or messy sources by itself. Useful results still depend on trusted business context.
Existing permissions still rule
Administrators choose the available connections and eligible roles. Queries enforce the connected account's table, row and column restrictions. Installing the agent does not automatically grant universal access.
Logs, permissions and sensitive outputs still need review. Local and cloud AI have different trade-offs. This product is useful precisely because it connects enterprise services, so governance is part of the feature.
It produces a deliverable, not just an answer
Users can refine an investigation, inspect supporting evidence and turn the result into an editable, shareable dashboard. The agent can also work with Omni, Oracle BI, Power BI, Sigma, Tableau and ThoughtSpot, then recommend or perform approved follow-up actions through connected tools.
Combining analysis, dashboards and actions makes it a workflow rather than a database chatbot. It follows the pattern in our guide to what AI agents are and how they work: context, tools and a reviewable sequence built around a task.
How to enable it and what to verify
Data is listed as a plugin in ChatGPT Work. An administrator installs it, enables the relevant data-source plugins and controls access. Employees then start a conversation with @Data. Effective availability depends on each workspace's configuration.
For consequential decisions, ask for evidence, verify definitions and compare a sample against a known report. Removing SQL from the interaction does not remove bad-data risk. The gain is faster investigation without hiding the trail. OpenAI says nearly all of its product team and more than two-thirds of its go-to-market organization already use data agents, evidence of internal maturity rather than independent proof of accuracy.
The conversation starts here
Sign in with a supporter account to comment. Sign in
Nobody has commented yet. Want to go first?