OpenAI released a Data agent in ChatGPT Work on September 10, 2026 — the same day it opened its Agents API to public beta for developers. The Data agent connects directly to company data sources, investigates what changed in business metrics, and builds interactive dashboards that employees can share — all from a plain English conversation.

This is not a chatbot that talks about data. It is an agent that goes into your data warehouse, runs the analysis, and hands you a working dashboard. That is a meaningful shift.

What the Data Agent Can Actually Do

The agent handles the kinds of questions that have historically required a data analyst or hours in a BI tool:

• Why did sales drop in Q3?
• Which customer segments are churning fastest?
• Where is spending increasing month-over-month?

Users direct and refine the analysis in a single conversation, without writing SQL queries or learning a separate tool. The agent investigates the data, builds the answer, and creates a shareable dashboard.

The Data agent connects to a broad range of data sources out of the box, including Amazon Redshift, Google BigQuery, Snowflake, Databricks, ClickHouse, MongoDB, and Datadog. It can also pull in files and documents from Google Drive and SharePoint to combine structured and unstructured data in the same analysis.

For dashboard delivery, it integrates with Power BI, Tableau, Sigma, ThoughtSpot, Omni, and Oracle BI — the tools most enterprise teams already use. An analysis built in ChatGPT Work can be pushed directly into a team’s existing BI environment.

What This Means for Business

This changes the conversation around data access inside organisations.

Right now, most business teams are data-adjacent. They know the questions they want answered but rely on data teams or pre-built reports to get them. The Data agent closes that gap. A finance manager can ask why budget variance is higher than forecast and get a proper breakdown without filing a request with the analytics team.

That has implications in both directions.

For business leaders, it means faster decisions. The bottleneck between a question and an answer shrinks from days to minutes. Teams that would not have used data now can, which means data starts influencing more of the day-to-day decisions that drive results.

For data professionals, the role shifts. Less time generating standard reports that anyone with the right tool could now produce themselves. More time on the work that requires genuine expertise: data architecture, model validation, governance, and the complex analyses that cannot be summarised in a single question.

The teams that adapt fastest will be the ones who understand both sides — the AI tooling and the underlying data fundamentals that make analyses trustworthy.

The Catch: Bad Data Still Produces Bad Answers

The Data agent connects to whatever data you give it. If your Snowflake instance has inconsistent schema definitions, outdated records, or missing joins between key tables, the agent’s analysis will reflect that. The garbage-in, garbage-out problem does not disappear when the interface is conversational.

This is actually an argument for investing in data literacy and data quality before deploying tools like this at scale. An organisation where everyone understands what the data means and how it is structured will get far more value from the Data agent than one that hands it a poorly maintained data warehouse.

Data cleaning, governance, and schema design are not less important in an AI-first analytics environment. If anything, they matter more, because now many more people are running queries and trusting the results.

The Competitive Landscape

Microsoft has been building toward this with Copilot in Power BI for the past two years. Google has been pushing similar natural language analytics through BigQuery and Looker. What OpenAI is doing here is putting that capability inside the conversational workspace that many teams are already using for other tasks — which lowers the barrier to adoption considerably.

The integration with Power BI and Tableau specifically matters because those tools dominate enterprise reporting. OpenAI is not asking companies to move to a new analytics platform. It is adding an intelligence layer on top of the tools they already have.

What Enterprise DNA Is Watching

For organisations building data skills alongside AI tools, this development reinforces a few things:

The fundamentals remain critical. SQL, data modelling, and understanding how to structure data for analysis are still the foundation. AI tools amplify those skills; they do not replace them.

Teams with strong data literacy will extract more value. Knowing whether an AI-generated analysis is correct requires understanding the domain and the data. That judgment cannot be delegated to the tool itself.

The use cases that benefit most from natural language analytics are operational questions with good underlying data. Revenue analysis, customer segmentation, support ticket trends, pipeline reporting. These are the areas worth testing the Data agent against.

If your team is already working in Power BI or Tableau, it is worth running pilots with the Data agent on your most common analytical requests and measuring whether the quality and speed of outputs improves. The tool is available through ChatGPT Work as a plugin in the Plugins directory.

---

Want your team to get more from AI-powered analytics? Enterprise DNA’s training programs cover Power BI, Python, SQL, and AI-assisted data workflows — the skills that make tools like OpenAI’s Data Agent actually useful. Explore learning plans for individuals and teams.