AI Assistants
BI Pixie records how people actually use your Power BI reports, but that evidence is only useful if the people and tools making decisions can reach it. Reading it through the dashboard means opening a report and knowing which page to look at, and questions that fall outside the prepared visuals ("which reports has this team stopped opening since the redesign?") require someone to build a new view first. AI assistants remove that step: they answer engagement questions in plain language, and they can bring that context into work already under way, such as deciding what to retire, where to invest, or which audience to survey.
The AI Assistants page connects your AI tools to the BI Pixie semantic model, so that people and agents alike can explore report engagement, adoption, and user satisfaction, grounded in the same data as your BI Pixie Dashboard. There are two options:
- Fabric Data Agent: a native, zero-install chat agent for everyone in your workspace. Ask in plain language inside Microsoft Fabric and Microsoft 365 Copilot. Best suited to business users.
- BI Pixie MCP Server: connect your own AI tools (Claude, GitHub Copilot, Codex, Azure AI Foundry) for exact, repeatable, programmatic access. Best suited to developers, analysts, and automation.
Open AI Assistants from the left sidebar of the BI Pixie portal at app.bipixie.com. You can use either option, or both.
Compare the two
| Fabric Data Agent | BI Pixie MCP Server | |
|---|---|---|
| Where it runs | Inside Microsoft Fabric | Local in VS Code, or remote (hosted) |
| Ask from | Copilot in Power BI, Microsoft 365, and Copilot Studio, plus Claude, GitHub Copilot, and Codex through the Data Agent's preview MCP server | Claude, GitHub Copilot, Codex |
| Best for | Business users, developers, and AI coding tools | Developers and AI coding tools |
| Requires | BI Pixie installed in your own Fabric workspace, on a paid Fabric capacity | Any BI Pixie Dashboard deployment |
Add a Fabric Data Agent
A Fabric Data Agent runs inside Microsoft Fabric, grounded on the BI Pixie semantic model in your own workspace. Once it exists, anyone with access to that workspace can ask questions about report usage in plain language, with no setup of their own.
Before you start
- BI Pixie must be installed in your own Fabric workspace. If you are still using the Power BI dashboard app, switch on the Overview page first, as described in Set up Dashboard. Until then, the AI Assistants page shows the Data Agent option as unavailable and explains what unlocks it.
- That workspace must sit on a paid Fabric capacity. Trial capacities do not support data agents, and creation will be refused.
- Your tenant's Copilot and Azure OpenAI setting must be enabled by a Fabric administrator.
- You need permission to create items in the workspace. Creating a data agent is a Fabric permission that not every account holds.
Create it
- Open AI Assistants from the left sidebar.
- On the Add an AI Data Agent card, select Add data agent.
- Approve the permission request if you are prompted. BI Pixie creates the agent under your own identity, grounded on your BI Pixie semantic model.
- When it is ready, the card changes to Open data agent, which opens it in Microsoft Fabric in a new browser tab.
Creating the agent is safe to repeat. If one already exists, BI Pixie reuses it rather than creating a duplicate.
Publish it
A new data agent is created in draft. Open it in Fabric and publish it once, which is what makes it available to other people in the workspace and to Copilot. After that, anyone with workspace access can start asking questions.
Use it as your AI provider
The data agent has a second role: it can act as the AI provider for AI Readiness, which is the component that writes the descriptions, AI Instructions, folder names, and benchmark question wordings BI Pixie proposes for your semantic models. Choosing it means there is no API key to paste and no separate account to manage, because inference runs on your own Fabric capacity.
Once the agent exists, its card offers Use as AI provider. Selecting it opens the same setup dialog the AI Readiness page uses, with the BI Pixie data agent already presented as the recommended option, so that you can create the agent and connect it without leaving this page. Save, and the AI Readiness provider status row shows it as connected.
This action appears only for admins, because the AI provider is a shared account setting rather than a personal one. You can change or remove the provider later from the Account page, on the AI tab. To use your own model instead, select Bring your own AI provider in the same dialog and connect Azure AI Foundry, OpenAI, or Anthropic.
If it will not create
- "You need permission to create a Data Agent in this workspace": ask a workspace admin to grant you contributor access, or to create the agent on your behalf.
- Any other failure: the cause is most often a trial capacity or a tenant Copilot setting that is switched off. Check both with your Fabric administrator, then try again.
Use the BI Pixie MCP Server
The BI Pixie MCP Server lets AI tools you already use query your BI Pixie semantic model directly. It is strictly read-only, so it can never change, delete, or refresh your data.
The quickest way to connect is from the portal:
- Open AI Assistants from the left sidebar.
- On the Use BI Pixie MCP Server card, select Install in VS Code. The link is already populated with your workspace and semantic model, so no manual configuration is required.
- Run a one-time
az loginso that the server can sign in as you.
To set it up by hand, or to connect a tool other than VS Code, see the MCP Server guide, which covers the full configuration, the available tools, and troubleshooting.
The install link carries your workspace and tenant IDs, never a password. You sign in once with your own Microsoft account, and the server only ever sees the data your account can already see.
Related
- BI Pixie MCP Server: full setup, tool reference, and troubleshooting.
- BI Pixie Semantic Model: what both options query.
- AI Readiness: make your own semantic models easier for agents to answer from.