Prophecy has launched DataGPT, a service for building and running data workflows inside ChatGPT, Claude and Copilot. Its central promise is more than asking a chatbot to analyse a spreadsheet: users can inspect the steps behind an answer, change them and approve a workflow before it runs again.
Watch Desk analysis
What happened
DataGPT can connect a messy Excel file with records in sources such as SharePoint or Snowflake, then produce a workflow for tasks such as reconciling transactions or applying a pricing rule. SiliconANGLE reports that each result comes with a visual map of the steps used. Workflows are backed by open code in Git, and approved runs can be scheduled on Snowflake, Databricks or Google BigQuery.
A person must review a workflow before it is scheduled. Prophecy says reruns follow the saved code rather than calling the AI model again, so they incur no additional AI cost. The company also says uploaded files pass through to an organisation’s own data platform rather than being stored inside the chat.
Why it matters
The product targets a familiar workplace workaround: upload a spreadsheet to a chatbot, accept the answer, and hope the logic survives the end of the conversation. DataGPT aims to turn that one-off exchange into a visible, editable process that can be checked and run again.
That could make AI-assisted analysis easier to govern, particularly for teams whose analysts do not read code. The practical difference is a workflow that can be inspected and approved, rather than a number arriving with a confident shrug attached.
Our read
The appealing idea is not that a chatbot can produce another spreadsheet answer. It is that the work behind the answer becomes something a colleague can examine, correct and repeat. That is a better starting point for business use than treating a chat window as an audit trail.
The pitch still needs to meet ordinary workplace tests: whether the workflow is understandable, whether the connections fit a company’s data rules, and whether the saved process behaves as advertised. But Prophecy is addressing a real gap between conversational AI and work people need to reproduce.
What to watch
- Whether analysts can reliably inspect and correct generated workflow steps.
- How organisations manage access to data sources and approved AI tools.
- Whether scheduled workflows deliver repeatable results across real business tasks.
Discussion spark: Would you trust an AI-generated data workflow once you can inspect and approve its code, or should someone with technical expertise sign off before it runs?
Sources and evidence
- Exclusive: Prophecy launches DataGPT to run data workflows inside ChatGPT, Claude and Copilot – SiliconANGLE (28 October 2026, 09:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.