ChatGPT Work gets Data agent for business analytics

What's new

On September 10, 2026, OpenAI introduced a Data agent in ChatGPT Work. It is a plugin (an add-on) that connects to a company's approved data sources and can answer questions like "why did sales drop?", "where are expenses rising?" or "which large customers are at risk of leaving?" The answer comes as an analysis and an interactive dashboard, a visual overview with figures and charts that you can share with colleagues.

The agent can pull data from sources including Amazon Redshift, Google BigQuery, Databricks, MongoDB, Snowflake and ClickHouse. It can also use files from Google Drive and SharePoint. You call on it with @Data, or ChatGPT can start using it on its own.

What's smart about it

You don't have to write queries (code that retrieves data) or learn a new analytics program. You type what you want to know, and can then refine the analysis in the same conversation.

The agent also tries to understand the company's own language: how a metric is defined and how tables relate to each other. To do that, it draws on sources such as dbt, GitHub, Snowflake Horizon, Databricks Genie Ontology and existing BI dashboards (BI stands for business intelligence). Dashboards can be connected to Power BI, Tableau, Sigma, ThoughtSpot, Omni and Oracle BI, so it complements those tools rather than replacing them.

Security has been built in: Queries follow the connected user's existing permissions, including at the table, row and column level. In other words, you can't see more through the chat than you are already allowed to see. Administrators can enable the agent for selected roles or groups.

Cheaper or better?

OpenAI has not disclosed a separate price for the Data agent. According to the coverage, access depends on an existing ChatGPT Work or Codex subscription with the plugin enabled. Monthly prices and information on availability in Denmark and the EU were not found in the sources.

A direct comparison with competitors' solutions has not been provided either, and there are no benchmarks. OpenAI itself says that nearly the entire product team and more than two thirds of the sales and marketing organization already use data agents internally. Among the early testers are NTT Data, Thermo Fisher and ServiceTitan. They use it to analyze sales and expenses, find errors in reports and choose which opportunities to pursue.

One important reality check: Conversational analytics does not fix bad data. If a company's numbers are messy, or two departments calculate metrics differently, the agent won't solve that on its own.

What it's good for

It suits companies that have already gathered their data in a data warehouse, and where ordinary employees in sales, finance or customer service need quick answers without waiting for a data analyst. Typical tasks include explaining swings in sales, tracking rising costs, finding errors in reports and building a dashboard for a meeting.

It is less relevant for small businesses without their own databases. And remember to double-check important figures before using them for major decisions.