September 14, 2026
On September 10, 2026, OpenAI introduced a Data agent for ChatGPT Work: a plugin that connects to a company's approved data sources, investigates what changed in a business metric, and builds an interactive dashboard employees can share, all through plain conversation. As Unite.AI reported, there's no query writing and no separate analytics tool to learn -- you describe what you want to know, and the agent goes and figures it out. OpenAI cited feedback from NTT DATA Group as an early proof point: according to that feedback, many non-engineers, particularly in sales and corporate functions, have been able to build and update their own dashboards using plain language, without waiting on a data or BI team. The Data agent connects to major warehouses and data platforms including Amazon Redshift, Snowflake, Google BigQuery, Databricks, and MongoDB, and it works alongside established BI tools like Power BI, Tableau, Sigma, ThoughtSpot, Omni, and Oracle BI. It's live now, listed under 'Data' in the ChatGPT Work Plugins directory, and admins can install it or roll it out to teams directly from Workspace settings.
We're not going to pretend this is a neutral news item for us. The premise behind the Data agent -- that a non-technical person should be able to describe what they want and get working output, no query language, no separate tool to learn -- is the exact bet ViibeStack is built on. We built our whole AI App Builder around the idea that plain-language description should be the interface for building things, not just for asking questions about things someone else already built. So it's a genuinely good sign for the whole category that OpenAI shipped a real, working version of this idea inside actual enterprise workflows, not a demo. The NTT DATA Group feedback OpenAI cited is the important part: people in sales and corporate functions -- not analysts, not engineers -- building and maintaining their own dashboards. That's the outcome every no-code and AI-builder vendor claims to want. Credit where it's due: this is a large, credible company putting plain-language, no-code interaction directly into the flow of enterprise data work, and that validates the direction the whole industry has been betting on, us included.
Here's the part worth sitting with. To get a dashboard from one plain-English sentence using the Data agent, a company needs a ChatGPT Work seat, plus the Data agent plugin itself, plus an underlying paid data warehouse or BI product for the agent to actually query -- Snowflake, Redshift, BigQuery, Databricks, or an existing Tableau, Power BI, Sigma, ThoughtSpot, Omni, or Oracle BI deployment. The plain-language layer is new. The stack underneath it, and the subscriptions that come with that stack, are exactly what they were before. That's not a criticism of OpenAI's execution -- the Data agent does what it says it does, and integrating with that many established platforms is real engineering work. But it means the Data agent is a front door bolted onto an existing structure, not a replacement for the structure. You still need to have provisioned a warehouse, still need to have someone maintaining a BI layer or at least a connection to one, and still need to be paying for all of it, before a salesperson can type a sentence and get an answer. The convenience is real. The cost and complexity underneath it didn't go anywhere.
If OpenAI has just proven that plain-language description is the right way for a non-technical employee to get analysis done, the conclusion shouldn't be that every business now needs to assemble a warehouse, a BI product, and a chat plugin on top of both just to answer 'what changed in this number.' That's three separate systems and three separate bills to answer one question. The better conclusion is that the reporting and dashboards a business needs should be built directly into the tool that's already tracking that data -- so there's no pipeline to assemble in the first place. This is the difference between our Analytics & Reporting and a bolt-on chat layer: when your CRM, your project tracker, and your finance records live in one system, asking 'what changed in this number' in plain language doesn't require a warehouse-to-BI-to-plugin relay -- the agent building your dashboard already has direct access to the source of truth, because it's the same platform. We've made this same argument before about Salesforce naming its agents as if a friendlier interface changes what's required underneath it. A plain-language front end is genuinely valuable. It's just not a substitute for owning the data model it's querying, and it's worth checking our buy vs. build vs. ViibeStack comparison if you're weighing whether to add one more plugin to an existing stack or consolidate the stack itself.
If you already have Snowflake or Redshift and an established BI tool, and you're paying for ChatGPT Work seats anyway, the Data agent is a low-risk way to get more people asking questions of data that's already being managed well. That's a legitimate use case, and it's worth trying. But if you're a smaller or mid-sized business looking at this announcement and wondering whether you now need a warehouse, a BI subscription, and a chat plugin just to get a dashboard, that's the wrong takeaway. The right takeaway is that plain language should be how you interact with the system tracking your data, period -- not an extra layer purchased on top of two other systems you also had to buy and integrate.
Sources