OpenAI Data Agent
by OpenAI
Natural-language data analysis and dashboards over your warehouse, inside ChatGPT Work
OpenAI Data Agent is a plugin for ChatGPT Work that connects to approved company data sources such as Snowflake, BigQuery and Databricks, investigates why business metrics changed, and builds shareable interactive dashboards. It is aimed at business teams who want answers without writing SQL or waiting on an analytics queue.
OpenAI Data Agent, launched on September 10, 2026, is a Data plugin for ChatGPT Work, OpenAI's agentic workspace that shipped in July 2026, and is also available in Codex. Workspace administrators enable it under Workspace settings, choose whether it is available or pre-installed, and switch on the data-source plugins users may reach; users then call it with @data in Work mode. It connects to Amazon Redshift, Google BigQuery, ClickHouse, Databricks, MongoDB, Snowflake and Datadog, reads files from Google Drive and SharePoint, and draws business definitions from dbt, Databricks Genie Ontology, Snowflake Horizon, GitHub and existing BI dashboards. Given a question such as why sales slowed or which renewals are at risk, it investigates metric changes, returns findings with reviewable sources, builds interactive dashboards teammates can edit, share and refresh, recommends next steps, and executes user-approved actions through connected tools such as Slack or email. It also works alongside Omni, Oracle BI, Power BI, Sigma, Tableau and ThoughtSpot. Queries run under the connected account's existing table-, row- and column-level permissions, though one analysis notes that published dashboards copy analyzed data and so no longer inherit source permissions. Alpha customers named in coverage include NTT DATA, Thermo Fisher, ServiceTitan, Zipline, Empower, micro1 and CookUnity. OpenAI says the internal agent it grew from serves more than 3,500 of its own users across 70,000 datasets. It enters a crowded field of natural-language analytics from Databricks, Snowflake, Microsoft, Google, Tableau and ThoughtSpot, and OpenAI has not published accuracy benchmarks, latency, warehouse compute costs or separate pricing; access is bundled into ChatGPT plans that include Work.
A head of data or analytics at a company already standardized on ChatGPT Business or Enterprise, with a governed warehouse and semantic layer, who wants to shrink the ad-hoc dashboard request queue.
Business users get cited answers and editable dashboards from warehouse data in one conversation, under the permissions they already have.
At a Glance
- Category
- Business Intelligence
- Pricing
- Subscription
- Target Market
- CIOs, CTOs, Heads of Data and Analytics, Business Operations Leaders, Finance and Revenue Operations Teams
- Deployment
- Cloud-only
Key Features
- ✓Metric investigation
Diagnoses why a business metric changed across periods and returns evidence-backed findings with sources users can review.
- ✓Warehouse and file connectors
Queries Redshift, BigQuery, ClickHouse, Databricks, MongoDB, Snowflake and Datadog plus Google Drive and SharePoint files in one conversation.
- ✓Semantic context
Uses business definitions from dbt, Databricks Genie Ontology, Snowflake Horizon, GitHub and existing BI dashboards to interpret company-specific metrics.
- ✓Interactive dashboards
Turns an analysis into a dashboard teammates can edit, share and refresh, with optional brand guidelines applied to outputs.
- ✓Permission-aware queries
Runs every query under the connected account's existing table-, row- and column-level permissions, with admins controlling which sources are available.
- ✓Approved follow-up actions
Recommends next steps and, with user approval, shares findings via Slack or email or acts in other connected tools.
Capabilities
Use Cases
- •Sales slowdown diagnosis
A revenue leader asks why pipeline conversion fell last quarter and gets a cited breakdown by segment plus a shareable dashboard.
- •Renewal risk review
Customer success asks which issues threaten renewals among the largest accounts, combining warehouse usage data with support records.
- •Self-serve KPI dashboards
Non-engineering staff build and refresh departmental dashboards in plain language instead of filing requests with a central BI team.
- •Spend monitoring
Finance asks where spending is rising across cost centers and receives a breakdown it can validate before sharing with leadership.
- •Leadership readouts
An operations team drafts a KPI framework and leadership summary from warehouse metrics and SharePoint documents in one session.
Ideal For
Best For
- ✓Explaining week-over-week or quarter-over-quarter metric changes without writing SQL
- ✓Letting non-engineers build and refresh shareable KPI dashboards
- ✓Companies already on ChatGPT Business or Enterprise with Snowflake, BigQuery, Databricks or Redshift
- ✓Teams with mature dbt or warehouse semantic layers that encode trusted metric definitions
- ✓Leadership readouts that combine warehouse data with Google Drive or SharePoint documents
Not Ideal For
- ✗Regulated or high-consequence reporting, since OpenAI has published no accuracy benchmark and independent analysts advise validating outputs against trusted reports
- ✗Organizations without metric ownership or a semantic layer, where natural-language questions produce inconsistent definitions of revenue or retention
- ✗Teams needing strict control over who sees data after sharing, because published dashboards copy data outside source permissions
- ✗Companies not on ChatGPT, which have no standalone way to buy the agent
Deployment
Market Analysis
Pros
- ✓Broad connector coverage across major warehouses, files and BI tools from day one
- ✓Enforces existing table-, row- and column-level permissions on queries
- ✓No new seat to buy for organizations already paying for ChatGPT Business or Enterprise
- ✓Alpha customers such as NTT DATA report non-engineers building dashboards in plain language
Cons
- ✗OpenAI has not released accuracy or retrieval benchmarks, so buyers cannot quantify how often answers are wrong
- ✗Published dashboards copy analyzed data and no longer inherit source-system permissions
- ✗Quality depends on a maintained semantic layer and metric ownership; analysts warn it moves definition and validation work upstream rather than removing it
- ✗Latency, warehouse compute costs, SLAs and dashboard approval workflows are undisclosed
Pricing
ChatGPT Business
From $20/user/mo (annual); $25/user/mo monthly
- ✓Includes ChatGPT Work
- ✓Data plugin enabled by workspace admin
- ✓2-user minimum
ChatGPT Enterprise
Contact for pricing
- ✓Includes ChatGPT Work
- ✓Admin-managed data-source connections and roles
- ✓Sales-negotiated contract
OpenAI has not published separate pricing for the Data agent; it is bundled into ChatGPT plans that include ChatGPT Work. Business costs $25 per user per month billed monthly or $20 annually with a two-user minimum; Enterprise is sales-negotiated. XDA reports Plus ($20/month) as the minimum individual plan, and agent runs draw down the plan's usage allowance much faster than chat. Warehouse query costs incurred on Snowflake, BigQuery or Databricks are billed by those vendors and OpenAI has not disclosed typical consumption.
Security & Compliance
Sources
This page was written from 9 sources, 9 on domains other than openai.com.
- 1.community.openai.com — 1396488
- 2.unite.ai — openai introduces data agent in chatgpt work to analyze comp
- 3.xda-developers.com — openai chatgpt data agent announcement
- 4.aicybr.com — openai data agent chatgpt work bi data sources
- 5.aiagentslibrary.com — chatgpt data agent
- 6.releasebot.io — openai
- 7.allweatherfinance.com — openai launched a data intelligence agent that supposedly tr
- 8.digidai.github.io — chatgpt work data agent dashboard labor
- 9.usecarly.com — chatgpt work pricing
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
Subscribe