OpenAI launches a data agent for ChatGPT Work

OpenAI has introduced a data agent for ChatGPT Work that can analyze company information, create interactive dashboards, and recommend actions using natural language. The goal is to help more people answer business questions without writing queries or learning another analytics tool.

From data to answers

Questions like why sales fell, where spending is increasing, or which customers are at risk of not renewing often depend on reports prepared by other teams. With the new agent, you can ask the question directly and continue investigating in the same conversation.

The agent can connect to company-approved sources, including:

• Amazon Redshift, Google BigQuery, ClickHouse, and Snowflake
• Datadog, Databricks, and MongoDB
• Google Drive and SharePoint for bringing in files and documents
• Business intelligence tools such as Power BI, Tableau, Oracle BI, Sigma, Omni, and ThoughtSpot

It does more than read tables. It also uses each organization’s business definitions, relationships between data, and custom calculations. That context can come from semantic layers and sources such as dbt, GitHub, Snowflake Horizon, Databricks Genie Ontology, and BI dashboards.

Dashboards, evidence, and next steps

After analyzing the data, the agent can turn the results into an interactive dashboard with built-in visualizations. Your team can edit, share, and update it as needed, while adapting it to the company’s visual guidelines.

You can review each finding to understand what evidence supports it. You can also ask follow-up questions to compare periods, investigate the causes of a change, or check whether a conclusion holds up.

The agent can go beyond analysis: it recommends next steps, identifies the people who should be involved, and, when you approve, shares results through Slack or email or takes action through connected tools.

The company remains in control

Administrators decide which data connections are available and which teams can use them. Queries respect the existing permissions of the connected account, including restrictions on tables, rows, and columns.

This matters because a chatbot with access to corporate data should not become a way to bypass internal controls. The agent’s usefulness depends as much on the quality of the definitions, permissions, and safeguards configured by the organization as it does on the quality of its answers.

OpenAI says that nearly all of its product team and more than two-thirds of its commercial organization use data agents in ChatGPT Work. Companies including NTT Data, Thermo Fisher, and ServicePiston are participating in the Alpha program and using them to analyze sales and spending, find errors in reports, and decide which opportunities to prioritize or how to allocate staff.

How to use it

The agent appears as Data in the ChatGPT Work plugin directory. An administrator must install or enable it from the workspace settings and configure the required data connections.

Once connected, you can start a conversation with @Data. For example, you can ask: Diagnose why weekly active users changed last week, so it can identify possible causes, compare them with earlier periods, and suggest what to review next.

What this changes for you depends on where you work. If your company connects its systems and defines its metrics properly, questions that once required waiting for a report could be answered in a conversation. The key things to watch are whether the answers include evidence, respect permissions, and do not replace human review when a decision affects customers, money, or staff.