For years, getting meaningful answers out of your company’s data meant either hiring a data analyst, learning SQL, or wrestling with tools like Tableau, Power BI, or Looker — software that’s powerful in the right hands and completely impenetrable in the wrong ones. OpenAI just decided that’s a problem worth solving directly. On September 10, 2026, the company officially introduced the Data agent inside ChatGPT Work, a feature that lets users connect their company’s data sources and build interactive dashboards using nothing but natural language. No SQL. No BI certification. Just questions.
Why This Moment Makes Sense
OpenAI didn’t arrive at this overnight. The groundwork goes back to ChatGPT’s early Code Interpreter feature — later renamed Advanced Data Analysis — which let users upload CSVs and ask questions about them. That was genuinely useful, but limited. You were working with static files, one conversation at a time, with no persistence and no live data.
The jump from “upload a spreadsheet” to “connect your company’s production database and build a shareable dashboard” is significant. It’s also a direct response to where the enterprise market has been heading. Every major cloud vendor — Microsoft, Google, Salesforce — has been bolting AI onto their existing BI products. Microsoft Copilot sits inside Power BI. Google’s Gemini is woven into Looker. But those are AI features added to BI tools. OpenAI is pitching something different: a BI capability built natively inside an AI product that tens of millions of people already use every day.
The timing also tracks with where OpenAI has been pushing ChatGPT Work as a platform. GPT-6 Astra gave the underlying model significantly stronger reasoning and tool-use capabilities. The Agents API opened the door for more autonomous, multi-step workflows. The Data agent is, in many ways, the first consumer-facing product that stitches those capabilities together into something a non-technical user can actually operate.
What the Data Agent Actually Does
Here’s the breakdown of what OpenAI is shipping:
- Live data connections: Users can connect to databases, cloud storage, spreadsheets, and third-party SaaS tools directly from ChatGPT Work. The agent maintains those connections rather than requiring repeated uploads.
- Natural language querying: Ask “What were our top five markets by revenue last quarter?” and the agent writes and executes the query behind the scenes, then surfaces the answer in plain English or as a visualization.
- Interactive dashboards: The agent can generate dashboards that update dynamically as the underlying data changes. These aren’t static exports — they’re live views you can share with colleagues.
- Insight surfacing: Beyond answering direct questions, the agent can proactively flag anomalies, trends, or patterns it identifies in your data — the kind of thing that currently requires a dedicated analyst or a scheduled report.
- Collaboration: Dashboards and analyses can be shared across a team within ChatGPT Work, with colleagues able to ask follow-up questions of their own without needing to rebuild the context.
The interface is conversational throughout. You’re not switching between a chat window and a separate dashboard builder — it’s all one surface. That design choice matters more than it might seem. Context collapse — the moment when you leave a conversation to go do something in another tool — is where productivity dies.
What Data Sources Are Supported?
OpenAI hasn’t published an exhaustive connector list yet, but the announcement references integration with common business data sources including SQL databases, Google Sheets, Excel, and popular SaaS platforms. Given OpenAI’s existing partnerships and the direction of their enterprise push, expect connections to Salesforce, HubSpot, and major cloud data warehouses like Snowflake and BigQuery to follow if they aren’t already available at launch. The breadth of connectors will ultimately determine how seriously large enterprises can adopt this without custom engineering work.
How Does It Compare to Existing BI Tools?
This is the honest part. Tableau and Power BI aren’t going anywhere tomorrow. They have years of enterprise trust, sophisticated visualization options, governance controls, and deeply embedded workflows. The Data agent isn’t trying to match them feature-for-feature — it’s trying to make the 80% of business questions that never get answered (because nobody had the time or skills to run a proper BI query) actually get answered.
Think about the mid-level marketing manager who wants to know if last week’s campaign moved the needle on trial signups. Today, that person either waits for an analyst, exports data manually, or gives up. With the Data agent, they ask the question directly. That’s the market OpenAI is actually targeting — not power users with custom dashboards, but the much larger population of people who are currently locked out of their own company’s data.
Who This Changes Things For
The implications cut differently depending on where you sit.
Small and Mid-Sized Businesses
This is probably where the impact lands hardest, fastest. A company with 50 employees that can’t justify a full-time data analyst now has something close to one available whenever they need it. The ability to connect live data and get dashboard-quality output from a conversation changes the economics of data-driven decision-making for businesses that were previously priced out of it. OpenAI’s affordability push has been making this segment a priority, and the Data agent fits squarely into that strategy.
Data Analysts and BI Teams
The more interesting question is what happens to people who currently spend their days answering one-off data requests. If the Data agent handles the “quick question” volume — and it likely will for many organizations — analysts get freed up for harder problems: model building, data infrastructure, strategic analysis that requires judgment rather than query execution. That’s the optimistic read. The pessimistic read is that some of those roles simply shrink. Both things can be true simultaneously.
BI Vendors
Tableau, Looker, and Power BI have meaningful moats: existing enterprise contracts, established governance frameworks, and integrations built over years. But they’ve also been selling the promise that AI will make their tools more accessible. OpenAI is now offering that accessibility as a standalone product, without requiring users to adopt an entirely new BI platform first. That’s a real competitive pressure, even if it doesn’t manifest in churn immediately.
Microsoft is in an interesting position here — they have Copilot in Power BI, but they’re also an OpenAI partner and investor. How they respond to a feature that directly overlaps with Power BI’s value proposition will be worth watching closely. I wouldn’t be surprised if we see Microsoft accelerate the Copilot integration timeline inside Power BI as a direct response.
What This Means for Enterprise Buyers
If your organization is already paying for ChatGPT Work — which sits at the higher end of OpenAI’s enterprise pricing — the Data agent appears to be included rather than a separate add-on. That matters for procurement conversations. The question IT and data governance teams will ask first is security: where does the data go when the agent queries it, what gets stored, and who has access? OpenAI has been building out its enterprise trust documentation, but expect detailed questions about data residency and compliance before any serious deployment in regulated industries. OpenAI’s work in sectors like financial services suggests they’re aware of these requirements, but the Data agent raises the stakes considerably since it involves live production data rather than one-time queries.
Key Takeaways
- The ChatGPT Data agent is live inside ChatGPT Work as of September 10, 2026, and lets users query live company data and build dashboards through natural language.
- It targets the large population of business users who currently can’t access their own company’s data without analyst support — not the power users already running Tableau.
- Connector breadth will determine enterprise viability. SQL databases, Google Sheets, and Excel are confirmed starting points.
- BI vendors face genuine competitive pressure, particularly in the SMB segment, though enterprise displacement will take longer given contract cycles and governance requirements.
- Security and data governance questions will be the primary friction point for enterprise adoption — watch for OpenAI to publish detailed compliance documentation quickly.
Frequently Asked Questions
What is the ChatGPT Data agent?
It’s a feature inside ChatGPT Work that allows users to connect company data sources and interact with that data through natural language. Users can ask questions, generate visualizations, and build shareable interactive dashboards without writing SQL or using traditional BI software.
Who is the Data agent designed for?
Primarily business users without deep technical or analytical backgrounds — the people who currently rely on a data team to answer routine questions. It’s also useful for small businesses that lack a dedicated analyst. It’s not designed to replace advanced BI workflows for power users, at least not yet.
How does it compare to Microsoft Power BI or Tableau?
Those tools offer deeper visualization customization, stronger governance controls, and more mature enterprise integrations built over many years. The Data agent’s advantage is accessibility and the fact that it lives inside a product people already use daily. It’s less about matching features and more about removing the barrier to entry for data-driven questions entirely.
Is the Data agent included in existing ChatGPT Work plans?
Based on OpenAI’s announcement, the Data agent is part of ChatGPT Work rather than a separate paid tier, meaning existing Work subscribers should have access. Organizations should verify current plan details directly with OpenAI, particularly as enterprise agreements may have different terms.
The Data agent is a clear sign that OpenAI views ChatGPT Work not as a chatbot with enterprise features bolted on, but as a genuine productivity platform — one that can sit at the center of how businesses operate day to day. Whether it gets there depends on execution: connector quality, security credibility, and whether the natural language querying actually holds up across messy real-world data. The opportunity is obvious. Delivering on it reliably is the harder part, and that’s where OpenAI’s next few product updates will tell the real story.
