Analysis Mode for Power BI

Ask your data

Ask questions about the data of a semantic model in your own words — Copilot answers by running real DAX queries against the model and shows every query it ran, so every figure can be verified. When an answer is a breakdown, one click turns it into a real analysis tab. This page explains how to use it, what it can and cannot do, and exactly which information leaves your tenant. It uses the same Azure OpenAI setup as Copilot analysis creation.

How it works

1
In an open analysis, choose Ask data in the side rail and ask in your own words: “which weeks have a negative net cash flow?”, “top customers this year, and how did they develop?”. Answers come in the language you ask in.
2
Copilot works out which queries answer the question, runs them against the semantic model and interprets the results. Every answer carries an expandable DAX queries section listing each query with its runtime and result size — the arithmetic is auditable, not a black box.
3
The open analysis is the default frame: its dimensions, filters and time windows are assumed unless your question says otherwise. Click ✦ on a cell or row of the grid to anchor the conversation to exactly that coordinate — follow-ups like “why is this one so high?” need no further explanation.
4
When an answer is a breakdown — groupings × figures — a quiet Adopt line appears beneath it: keep it as a new analysis tab or apply it to the analysis you have open. If the adoptable layout differs from the answer (say, a full list instead of a top 3), the difference is stated right there.

An administrator can switch the capability off at any time under Business Central's Copilot & agent capabilities — the pane then disappears for everyone.

What it can and cannot do

  • Every figure comes from a query. The AI decides which queries to run; the numbers themselves are computed by the Power BI engine, under the signed-in user's identity, with row-level security applied. Check the DAX queries section when a figure matters — that is what it is for.
  • Answers are AI-generated — verify results. The interpretation around the figures is a language model's text. The pane says so, and the shown queries are the way to verify.
  • Field names come from your model. Copilot works with the model's real tables, fields and measures. If a question cannot be answered from them, it says so rather than inventing a figure — and when a term is ambiguous, it names the reading it chose.
  • Complex questions take several queries. A multi-part question can run a handful of queries and take some seconds; the steps stay visible while it works.
  • Same question, different route. Language models are not deterministic; asking again can take a different (equally valid) query route to the same answer.

Your data

Ask your data runs against an Azure OpenAI resource in your own subscription, called server-side from Business Central. The key is stored in Business Central's Isolated Storage and never reaches the browser. No navida service is involved, and AI consumption is billed by Microsoft Azure on your resource. Unlike Copilot analysis creation, answering questions about data requires showing the AI data: the results of the queries it runs travel to your Azure OpenAI resource — and only there.

Sent to your Azure OpenAI resource

  • The structure of the semantic model: table, field and measure names, descriptions and relationships
  • Your questions, verbatim, and the layout of the open analysis (plus the anchored cell, if any)
  • The results of the DAX queries the chat runs, and lists of the values common attributes contain — so questions can be matched to your data's real spellings

The boundaries that hold

  • Every query runs as the signed-in user — row-level security limits what the chat can ever read
  • Data travels only between Business Central, your Power BI tenant and your own Azure resource — never to navida or any third party
  • Read-only: the chat can query the model, never change it or write anything back

Azure OpenAI is an enterprise service in your subscription: prompts and responses are not used to train models, and your Azure region and compliance boundary apply — see Architecture for the app's overall data flow.

Set-up in short

There is nothing separate to configure: the same Azure OpenAI resource that powers Copilot analysis creation powers Ask your data — endpoint, deployment and key on Power BI Analysis Setup, with Enable Copilot switched on. The step-by-step guide is on the Install page. Both features are included in the app licence — no separate plan.

Contact us

Need support? Use this contact link to send us an email.