Prophecy launched DataGPT on October 28, 2026, a tool that lets business data analysts do complete data work, from onboarding messy files to publishing finished reports, inside ChatGPT, Claude and Microsoft Copilot. Every result comes with a visual data workflow that can be inspected and validated, and approved work ships as repeatable code on the data platform the organisation already owns, running on Snowflake, Databricks or BigQuery. It is available today, with a free account option, and inside the Prophecy app.
The Problem DataGPT Targets
The problem is familiar to anyone who has pasted a spreadsheet into a chatbot. When analysts upload spreadsheets to reconcile a quarter of transactions, the AI writes and runs a script behind the scenes to produce a result, but because they cannot read that code, they cannot validate it, refine it or stand behind it. Once the chat session ends, the workflow vanishes with it, along with the domain knowledge supplied along the way, such as business rules, revenue definitions and approved exceptions.
That gap is growing as AI-generated data spreads. Prophecy cites a Gartner prediction that by 2028, half of organisations will adopt zero-trust data governance as unverified AI-generated data grows. The point is not that chatbots are bad at data work, but that a one-off answer is the wrong artefact for work that must be repeated, audited and defended. A finance report is not a conversation; it is a process, and a process needs to survive the session that created it.
Prophecy’s founder and chief executive Raj Bains framed the launch as closing a gap between coders and analysts. Coders got AI that collaborates with them from development to production, he said, while business data analysts got question answering and nothing else changed. DataGPT is meant to give analysts what coders got: the visual workflow behind every result, in a form they can verify and change, built on governed knowledge.
How DataGPT Works
In a single conversation, an analyst can upload a messy Excel file, check it against an audit list and data in SharePoint, Snowflake and Databricks, apply a pricing rule for discounted product bundles, review the results and schedule the workflow to rerun every month. DataGPT connects to files, SharePoint, data warehouses, ERP and CRM systems, and every workflow is backed by open code in Git, so the logic stays transparent and auditable.
The shift is from building to reviewing. With DataGPT, analysts no longer build every workflow from scratch; AI builds it, and they verify it visually instead, reviewing each step and its results on screen, making small refinements where needed and approving the final version. Approved workflows then rerun exactly the same way every time, with no new AI generation and no added AI cost, which matters both for consistency and for budgets.
A knowledge graph connects business context with data workflows, helping teams apply relevant definitions and business rules across their work. That is the mechanism behind the “compound knowledge” claim: once a team has defined how it treats a discounted bundle or an approved exception, that definition travels with future workflows instead of being re-explained in every chat.
Bringing Shadow AI Under Governance
Many business analysts already paste company spreadsheets into AI chat tools without IT’s knowledge or approval, one of the most common forms of shadow AI. With DataGPT, uploaded files go through Prophecy to the organisation’s own data platform, not into an AI chat, and data platform leaders and IT can see every workflow and govern it on the platform they already own, without adding another silo.
That governance pitch is likely to matter more to buyers than any single feature. IT leaders do not need another dashboard; they need a way to say yes to AI-assisted data work without losing control of where company data goes. By routing work through the existing platform and keeping the code open in Git, Prophecy gives IT both visibility and an audit trail, while analysts keep the conversational interface they already like. It is a compromise both sides can live with, provided the workflows genuinely are as inspectable as claimed.
The wider context is a wave of tools turning chatbots into work surfaces, from autonomous data analysts to analytics agents and agents inside industry ERP systems. DataGPT’s distinguishing bet is that verification, not generation, is the scarce resource for business data, and that the tool which makes verification easy will win the analyst’s desk.
Why It Matters
This matters because the bottleneck in enterprise data work has shifted from producing an answer to trusting one. Anyone can now generate a plausible report in a chat window; far fewer can prove every number in it, repeat it next month and explain it to an auditor. If DataGPT’s visual workflows deliver on that promise, it turns AI from a clever shortcut into a governed production process, and that is a much larger prize. The sceptical view is that visual inspection only helps if analysts actually inspect, and that a pretty workflow can launder a bad assumption as easily as a chat answer can; the test will be whether teams catch errors they would previously have missed.
Background
Prophecy, founded in 2017 and headquartered in Palo Alto, California, is the AI platform for business data work, with more than 130 employees and more than $100 million raised, and customers including SAP, HSBC, JP Morgan Chase and Pfizer.
Source: Prophecy announcement
0 Comments