# Prophecy DataGPT: Pricing and Data Workflows in Claude, ChatGPT and Copilot

Prophecy puts inspectable data pipelines inside AI conversations, but the client, pricing tier and review process still matter.

By Paras Madan | 2026-10-29

Prophecy launched DataGPT on October 28, letting business analysts prepare data and produce reports through Claude, ChatGPT and Microsoft Copilot. Each result includes a visual workflow that can be inspected and turned into repeatable code, giving teams something more durable than a one-off answer.

The useful question is what happens after the first convincing chart. Can someone check the joins, correct the business rules and run the same work next month without starting another conversation?

## What Prophecy DataGPT does

The [launch announcement](https://www.globenewswire.com/news-release/2026/10/28/3389911/0/en/prophecy-launches-datagpt-bringing-enterprise-data-workflows-to-claude-chatgpt-and-copilot.html) describes a workflow covering data onboarding, preparation, analysis and output. Analysts describe an outcome in their existing AI assistant, while Prophecy produces the underlying data work.

According to the [DataGPT product page](https://www.prophecy.ai/datagpt), approved workflows can run on embedded DuckDB or supported cloud platforms, including Snowflake, Databricks and BigQuery. Teams can inspect transformations and underlying code, retain approved business definitions and schedule recurring work.

For a monthly reconciliation, that means the useful artifact is the procedure for matching records and handling exceptions. The finished chart is only one output. A repeatable procedure gives the next reviewer a concrete place to look when totals change.

## DataGPT pricing and the free plan

Prophecy is offering a free starting point, but cloud execution and enterprise deployment are separate considerations. Its [current pricing](https://www.prophecy.ai/pricing) lists:

| Plan | Published price | Relevant allowance |
| --- | --- | --- |
| Starter | Free | 20 credits per month; embedded DuckDB processing |
| Professional | $150 per user per month, billed monthly | 50 credits per user per month; Databricks, Snowflake and BigQuery processing |
| Enterprise Express | $4,000 per month, billed annually | Up to 20 users; dedicated SaaS deployment |
| Enterprise | Custom | Larger enterprise deployments |

Professional also lists additional usage on a pay-as-you-go basis. These are Prophecy plan prices, not an all-inclusive estimate covering an external AI assistant and an organization's cloud infrastructure.

Prophecy says approved scheduled workflows rerun without prompting the model again. That can reduce repeated generation work, but **no new model generation does not mean no compute cost**. Teams still need to understand their execution platform and usage allowances.

## Claude, ChatGPT and Copilot are not identical clients

The [connection documentation](https://docs.prophecy.ai/data-analysis/getting-started/datagpt/connect-datagpt) describes a custom MCP connector authenticated with a Prophecy account. MCP is the protocol through which the assistant accesses Prophecy's tools.

Claude and ChatGPT can show interactive pipeline canvases, tables and dashboards inside the conversation. The documented Copilot Studio integration returns text and links; users open Prophecy to see the interactive views. That distinction matters when choosing where analysts will review generated work.

The connector uses each person's Prophecy access, even when an administrator adds it for the organization. Installing a shared connector does not give every employee the administrator's permissions.

## What still needs human review

Prophecy's [data-handling documentation](https://docs.prophecy.ai/data-analysis/getting-started/datagpt/permissions) says results can include sample rows in interactive views. It also describes confirmation steps for actions such as publishing projects and deleting files. Teams should inspect those data flows rather than assume that keeping execution in Prophecy means nothing sensitive appears in the assistant interface.

An inspectable workflow also remains a workflow someone must inspect. A wrong fiscal-calendar rule can be applied consistently every month. Reproducibility makes that mistake easier to find and correct; it does not establish that the initial business logic was right.

The strongest evaluation would compare a real recurring report against its existing approved result, then deliberately change an input or mapping. What matters is whether an analyst can explain the difference and safely reuse the corrected workflow.

I’m Paras. I work at the intersection of AI, engineering and distribution.

I help technical teams turn complex AI products and enterprise workflows into clear stories that developers and decision-makers can understand.

[Book a 30-minute call](https://cal.com/paras-madan/30min).