ChatGPT Work Data Agent Review 2026: Connect Business Data and Build AI Dashboards
OpenAI has launched a Data agent for ChatGPT Work that can investigate business questions, analyze governed company data, and build interactive dashboards through conversation. Announced on September 10,.
OpenAI has launched a Data agent for ChatGPT Work that can investigate business questions, analyze governed company data, and build interactive dashboards through conversation. Announced on September 10, 2026, the agent connects to databases, analytics platforms, business-intelligence tools, files, and semantic layers while enforcing the user’s existing access permissions.
The goal is not merely to generate a chart. The Data agent is designed to investigate why a metric changed, examine evidence, create a reusable dashboard, recommend actions, and continue refining the analysis in the same conversation.
What is the ChatGPT Work Data agent?
Data is an installable agent in the ChatGPT Work Plugins directory. After workspace administrators enable it and configure approved connections, employees can invoke @Data, ask a business question in plain language, and guide the analysis through follow-up prompts.
The agent can query structured data, combine it with files and documents, interpret company-specific metric definitions, and turn findings into interactive dashboards. It is aimed at widening access to analysis without discarding governance or evidence.
Supported data sources
OpenAI lists connections including Amazon Redshift, Google BigQuery, Snowflake, Databricks, ClickHouse, MongoDB, Datadog, Redis, Google Drive, SharePoint, and others. The agent can also use trusted business context from dbt, GitHub, Databricks Genie Ontology, Snowflake Horizon, and existing BI dashboards.
Why semantic context matters
A database column rarely explains what a company means by “active customer,” “qualified pipeline,” or “retention.” The Data agent can use semantic layers, business terms, custom calculations, and documented relationships so it interprets questions using the organization’s established definitions.
This does not guarantee correctness. Teams still need owners for metric definitions, tested joins, fresh data, and a process for resolving conflicts between sources.
From a question to an interactive dashboard
A user can ask why weekly active users declined, inspect the drivers, compare segments and prior periods, and request a dashboard summarizing the result. The dashboard can be edited, shared, refreshed, and styled using company brand guidance.
OpenAI highlights workflows for adoption and retention, business performance, operations, product growth, and financial analysis. The agent can also build and interact with dashboards in Power BI, Tableau, Sigma, ThoughtSpot, Omni, and Oracle BI.
“Diagnose why weekly active users changed last week. Identify likely drivers, compare with prior periods, show the evidence, flag data-quality concerns, and recommend the next checks.”
Permissions and enterprise controls
Workspace administrators decide which connections and roles are available. Queries enforce the connected user’s existing table, row, and column permissions. That is a crucial distinction: installing an agent should not automatically grant broader database access.
Organizations should still validate service-account scopes, dashboard-sharing rules, exported files, logs, and actions triggered through connected tools. Sensitive analyses may need extra approval and retention controls.
Analysis that can lead to action
After producing findings, the Data agent can suggest next steps and identify stakeholders. With approved connected tools, it can share results through Slack or email and carry out authorized follow-up actions. Human review is essential when the result affects customers, employees, spending, or operational systems.
How to install the Data agent
- Open the Plugins directory in ChatGPT Work.
- Find the plugin named Data.
- Install it or ask a workspace administrator to make it available.
- Enable and configure the required data-source plugins.
- Start a conversation with @Data and ask a scoped business question.
Who benefits most?
- Product and growth teams investigating adoption, conversion, and retention.
- Sales and finance teams analyzing pipeline, revenue, forecasts, and spending.
- Operations leaders monitoring service performance and exceptions.
- Analysts creating governed self-service workflows for nontechnical teams.
- Executives who need evidence-backed, refreshable performance views.
Limitations and risks
- Incorrect metric definitions or joins can produce polished but misleading answers.
- Access controls need testing across databases, files, dashboards, and outputs.
- Large investigations still require analytical judgment and domain context.
- Generated recommendations may confuse correlation with causation.
- Availability depends on ChatGPT Work, administrator approval, and compatible data connections.
Data agent vs a traditional BI dashboard
A traditional dashboard is excellent for repeatedly monitoring known metrics. The Data agent is more useful when the question is exploratory: what changed, why it changed, which evidence supports the conclusion, and what should be examined next. The strongest setup uses both, letting governed dashboards provide trusted definitions while the agent accelerates investigation and communication.
Verdict
The Data agent could make serious business analysis available to far more employees, but its real value depends on the foundation underneath it. Clean data, shared definitions, tested permissions, and reviewable evidence matter more than the conversational interface alone. Organizations with that foundation can use the agent to shorten the path from question to investigation, dashboard, and action.
Explore OpenAI’s official Data agent information
Source note: Based on OpenAI’s September 10, 2026 announcement. Connections, availability, and capabilities may change.