Data agent
Also known as: analytics agent, BI agent, data analysis agent
A data agent sits between a business user and their data stack. Instead of querying a BI tool directly or filing a ticket with a data analyst, a user asks a question in plain language: 'Show me last quarter's revenue by region compared to the same period last year.' The agent figures out which data sources to pull from, runs the query or analysis, and returns a chart, table, or narrative answer in the same interface.
OpenAI launched a Data agent inside ChatGPT Work in September 2026, letting businesses connect data sources, ask questions in natural language, and generate interactive dashboards. The release is part of a broader pattern: AI agents moving from standalone tools into the fabric of everyday work software, embedded alongside the spreadsheets, CRMs, and communication tools where decisions actually get made.
The practical benefit is the reduction of the analyst bottleneck for routine questions. Analysts can focus on the analysis that actually requires judgment while the agent handles the mechanical retrieval and formatting. The current limitation is trust: a data agent that silently uses the wrong date range, joins tables incorrectly, or samples data without disclosing it can produce plausible-looking answers that are wrong. Good data agent implementations surface the query they ran and the data they used so results can be spot-checked.