ChatGPT Work + Power BI: What the Integration Can (and Can't) Do in 2026
Partly — there’s no official one-click ChatGPT connector for Power BI, but Microsoft ships a remote Power BI MCP server (public preview) that its own docs walk you through registering with ChatGPT. Power BI isn’t in ChatGPT Work’s set of official connectors — those cover Microsoft 365, Google Drive, Slack, and Notion, and Microsoft steers BI work toward its own Copilot. What exists instead is a custom-MCP path: Microsoft’s remote server at https://api.fabric.microsoft.com/v1/mcp/powerbi lets ChatGPT ask questions of your semantic models in natural language, generating and running DAX under the hood. It’s read-only querying, it authenticates through a Microsoft Entra app you register yourself, and — like every MCP connection in ChatGPT — it only works inside a session you’re driving, in agent runs that are metered.
Here’s what that connection actually does, how to stand it up, where the ceiling is, and what to use if you want Power BI-adjacent work that runs on its own.
What ChatGPT Work can actually do with Power BI
Through Microsoft’s remote Power BI MCP server (the query path):
- Ask your semantic models questions in plain English. “What were last quarter’s sales by region in the Revenue model?” — ChatGPT calls the server, which generates and executes a DAX query using the same engine behind Copilot for Power BI, and returns the numbers in chat.
- Explore schema and relationships. The server is schema-aware, so it can learn a model’s tables, columns, and measures and answer “what tables are in this semantic model?” or “which measures reference Net Revenue?”
- Run ad-hoc analysis without building a report. Conversational slices and trends against a live model, no Power BI Desktop open.
- Respect your permissions. Queries run as the signed-in user under Fabric role-based access control, so ChatGPT only reaches models and workspaces that user can already open.
- Fold into an agent run. In ChatGPT Work (launched July 9, 2026), an agent can pull Power BI figures into a broader task — say, drafting a board deck that cites this quarter’s numbers alongside data from your other connectors — in one long, metered run you start.
There’s a second, different server worth naming so you don’t conflate them: the local Power BI modeling MCP server (microsoft/powerbi-modeling-mcp). It does write — creating tables, columns, measures, and relationships — but it’s a developer tool that runs locally over stdio for model authoring in VS Code or Power BI Desktop, not a hosted endpoint ChatGPT Work reaches into. The remote server, the one you’d wire to ChatGPT, is query-only.
How to set it up
This is an admin-gated, custom-MCP flow — not a checkbox in the connector gallery. Per Microsoft’s external-clients guide:
- Have your Power BI admin enable the tenant setting “Users can use the Power BI Model Context Protocol server endpoint (preview).”
- Register a Microsoft Entra app in the same tenant as your Power BI environment (single-tenant). Copy its Application (client) ID.
- Add the ChatGPT redirect URI (
https://chatgpt.com/connector/oauth/<random_chars>, exactly as ChatGPT gives it) as a Mobile and desktop redirect on the app. - Grant delegated Power BI Service permissions —
Dataset.Read.All,MLModel.Execute.All, andWorkspace.Read.Allonhttps://analysis.windows.net/powerbi/api. All read scopes. - In ChatGPT, go to Settings → Apps → Create App. Name it
Power BI, set the MCP Server URL tohttps://api.fabric.microsoft.com/v1/mcp/powerbi, choose OAuth, and under Advanced settings pick User-Defined OAuth Client and paste your Entra Application (client) ID. - Sign in and consent, then test with something read-only: “What tables are in semantic model [ID]?”
Creating custom MCP connectors in ChatGPT is a developer-mode capability that Enterprise/Edu admins gate by role, so on managed workspaces this is a coordinated setup between your Power BI admin and your ChatGPT admin, not a solo click.
The limits that matter
- It’s read-only. The remote server queries models and generates insights — it does not update reports, refresh datasets, publish, or write back through this path. The delegated scopes are all
Read/Execute, no write. - Preview, and it can change. Both the remote and local servers are public preview; Microsoft notes tool definitions, request formats, and response schemas may shift.
- No triggers. Nothing fires on a Power BI event. There’s no “when the nightly refresh finishes, send the summary” or “alert me when this KPI crosses a threshold.” ChatGPT touches Power BI only when you prompt it — a dashboard can move all weekend and nothing goes out.
- Session-bound and metered. Even inside a ChatGPT Work agent run, the work is a session you launch, and workspace-agent runs are token-metered against your plan’s allowance. It’s an errand you send, not a standing watch on your numbers.
- The follow-through stops at the chat. ChatGPT can pull the figure and draft the commentary; it won’t email the number to the exec team from your mailbox, post it to the finance channel, and log it to a tracker as one motion.
If you want Power BI-adjacent work that runs on its own: Carly
Reporting is trigger-shaped. The refresh lands overnight, a metric slips on a Tuesday, the weekly digest is due every Monday at 8am. The moment you want the response to happen on the event — the KPI summary built and sent the minute the dataset refreshes, an alert when revenue-to-plan drops below target, the board pack assembled and emailed before the meeting — you’ve crossed past what a chat session is for.
That’s where Carly fits. Carly is an AI executive assistant built to act on triggers, not just answer in a session:
- Fires on events and schedules, 24/7, in the cloud. When the Power BI dataset refreshes or a threshold trips, Carly pulls the figures, writes the summary, and sends it — while your laptop is closed.
- Actually sends and updates. Carly drafts and sends email across Gmail and Outlook, posts to Slack, updates records and tasks in your CRM, and records meetings — the follow-through that stops at the chat with ChatGPT.
- Builds the workflow by interviewing you. Tell Carly “every Monday at 8, pull the sales-by-region numbers, write a three-line summary, and email it to the leadership list” in plain English; it interviews you and builds it — no prompt engineering, no admin center.
- Connects to everything. 200+ native integrations across 40+ categories, plus any other tool via your own API key.
Power BI isn’t a native Carly integration, but its programmatic surface is the Power BI REST API with Microsoft Entra OAuth — so you connect it with your own credentials: paste your API key on carlyassistant.com/integrations and Carly can do whatever that access allows. AI agents start at $35/month, and steps in a workflow that don’t use AI run free and unlimited. See integrations.
ChatGPT Work vs Carly
| ChatGPT Work (Power BI MCP) | Carly | |
|---|---|---|
| Ask semantic models questions in plain English | Yes, in a session | Yes |
| Generate and run DAX queries | Yes (remote server) | Via the API |
| Write back / refresh / publish | No (read-only) | Via scoped API access |
| Reacts to a dataset refresh by itself | No | Yes, on the trigger |
| Alerts when a KPI crosses a threshold | No | Yes |
| Monday reporting digest, on schedule | No | Yes |
| Emails the numbers to the team | No (drafts only) | Yes (Gmail + Outlook) |
| Runs while your laptop is closed | No (session-bound, metered runs) | Yes (cloud, 24/7) |
| Setup | Entra app + custom MCP, admin-gated | Describe it in plain English |
| Pricing | Paid ChatGPT plan, metered agent runs | AI agents from $35/mo |
ChatGPT’s Power BI connection is an analyst you query in a chat. Carly is a teammate that reports on your numbers while you’re doing something else.
Frequently Asked Questions
Does ChatGPT Work with Power BI?
Partly. There’s no official Power BI entry in ChatGPT’s connector gallery — Power BI isn’t among ChatGPT Work’s built-in connectors (Microsoft 365, Google Drive, Slack, Notion). But Microsoft’s remote Power BI MCP server (public preview) is officially documented to register with ChatGPT via a custom MCP connector, letting you query your semantic models in natural language. It’s read-only.
How do I connect ChatGPT to Power BI?
Your Power BI admin enables the MCP tenant setting, then you register a Microsoft Entra app, add ChatGPT’s redirect URI, and grant delegated Power BI read scopes. In ChatGPT, go to Settings → Apps → Create App, set the MCP URL to https://api.fabric.microsoft.com/v1/mcp/powerbi, choose OAuth with a User-Defined OAuth Client, and paste your Entra client ID. Full steps are in Microsoft’s external-clients guide.
Can ChatGPT update or refresh Power BI reports?
Not through the remote MCP server — it’s read-only, scoped to Dataset.Read.All, MLModel.Execute.All, and Workspace.Read.All. Microsoft’s local modeling MCP server can write to semantic models, but it runs locally over stdio for developers in VS Code or Power BI Desktop, not as a hosted endpoint ChatGPT Work connects to.
Can ChatGPT send my Power BI report automatically every week?
No. MCP connections in ChatGPT respond inside a session you start, and agent runs are metered — nothing fires on a schedule or a dataset refresh. For “every Monday, pull the numbers and email the summary” or “alert me when a KPI slips,” you need a trigger-based assistant like Carly, which runs in the cloud around the clock and connects to Power BI through your own API credentials.
More: Best AI agents for productivity · ChatGPT MCP · Can ChatGPT send emails · Best AI tools for executives · ChatGPT + Salesforce
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
