The AI AdoptionFree AI pre-audit

Can an AI dashboard show
how your business is doing?

  1. The questionA plain question about the business’s own figures.
  2. Where the figures come fromClaude says which records it will read, and from which system.
  3. The dashboardEach chart comes from a query you can open and check.
  4. The answerBack in the chat, with the figures that explain it.

A demo screen from Anthropic’s announcement of 8 October 2026, with our four marks.

An AI dashboard is now something you ask for in plain words. You type a question about your business, and the AI shows the answer as charts built from your own systems. Claude Data, Copilot in Power BI and Gemini in Google Sheets all work this way. Before you rely on what it shows, check where the figures come from. Then check one figure yourself, and see who else can open the dashboard.

By The AI Adoption · Published on 22 November 2026 · Sources: Anthropic, Microsoft and Google, read on 22 November 2026

Where the figures come from

Every AI dashboard reads its figures from somewhere. In Claude Data, that is a data warehouse your company connects, such as BigQuery or Snowflake, or another app already connected to Claude, such as Salesforce. In Power BI, Copilot answers from the semantic model behind the report, which is the prepared version of your data that the report is built on.

Microsoft says what happens when that model is not ready. When a question is about the data in the model, Copilot uses the model. Otherwise it answers from the general knowledge of the AI model, and without preparation it can misread the data.

A Power BI report of apartment bookings with four key figures and two charts, and the Copilot pane open on the right, offering to explain the report page or prepare a summary for the team.
The Copilot pane beside a report, in Microsoft’s Power BI documentation.

Without preparation, Copilot can misinterpret the data and return generic or inaccurate results.

From Microsoft’s Copilot for Power BI overview

The same goes for a chart made from a table pasted into a chat, or typed into a sheet by hand. Google says a chart that Gemini generates in Google Sheets does not change when the original data changes. Only a chart built from a BigQuery Connected Sheet stays linked and updates when you refresh it.

In most businesses the figures already live in a few places: the accounting software, the point of sale, the stock system, the CRM. A dashboard you can rely on reads those systems directly. The figure on screen is then the one in your accounts, and nobody copies numbers from one place to another before the weekly meeting.

Three ways to build one

The three tools below each build a dashboard or a chart from a question. They differ in what they read, how you check a figure and what they need before you start.

Claude Data

Reads
A data warehouse, or an app connected to Claude
To check a figure
Each chart shows its query, and when its data was last refreshed
Needs
A paid Claude plan. On Team it is on by default

Copilot in Power BI

Reads
The semantic model behind the report
To check a figure
Ask about the report’s own data. Otherwise Copilot answers from general knowledge
Needs
A paid Fabric or Power BI Premium capacity. A Power BI Pro licence alone is not enough

Gemini in Google Sheets

Reads
The sheet and its tabs, and BigQuery through Connected Sheets
To check a figure
Open the analysis steps, or show the code behind the answer
Needs
An eligible Google Workspace or Google AI plan

In Google Sheets, you open Gemini from the side panel and ask for the chart you want. Gemini explains what it will do with the data before it draws the chart, and you preview it before inserting it. Inserted, the chart goes on a new tab with the data it was built from, so the figures stay with the chart.

Google is rolling out its newest model, Gemini 4 Hydrogen, to Gemini in Google Sheets for eligible business subscriptions since 12 November 2026. Our page on the Gemini 4 releases sets out who has it.

Gemini in Google Sheets, in Google’s announcement of 29 January 2025.

Checking a figure before you rely on it

The first figure to check is one you already know. Take a number you can read in your accounting software today, such as last month’s sales or last quarter’s VAT or GST, and ask the dashboard for it. If it gives the same figure, it reads the same source. If it does not, the difference tells you what it counted.

In Claude Data you open the chart’s query to see how Claude got the number, and tell Claude what to change if it is not counting what you meant. A data team can also approve official queries, which Anthropic calls certified queries. Charts built on them carry a badge, and the feature is still in beta.

  • Ask for a figure you knowLast month’s sales, from the accounting software
  • Compare the twoSame figure: same source. A different one: something is counted differently
  • Open how it was countedThe query behind the chart, or the analysis steps
  • Correct it onceTell the AI what to count, then ask again
On openingClaude Data refreshes a dashboard when someone opens it and the data is due, not on a timetable
FixedA Gemini chart made from sheet data does not follow later changes to that data
24 hoursCopilot in Power BI repeats its answer to the same question on an unchanged model within this window

Who sees what

A dashboard shows business figures, so it matters who can open it. By default, each tool works from what the person can already open. Live sharing in Claude Data is the exception.

Shared the usual way

  • Each person’s own connections load the data
  • Someone without access to a source does not see it

With live sharing in beta

  • Everyone you share with sees results from your connection
  • Even without access to that source

Anthropic’s advice is to check what a dashboard shows before you share it live. A dashboard of margins by job built on the owner’s access to the accounts would show those margins to everyone it is shared with.

Google says Gemini only retrieves the content in Workspace that the person has access to, and that company data is not used to train its models. Whatever a person cannot open in Drive, Gemini cannot open for them.

In Power BI, where the data is processed matters too. If your Fabric capacity sits outside the United States or the EU data boundary, in Australia for example, Copilot stays off until your administrator allows the data to be processed outside your region. Our guide on what you can safely share with AI lists the questions to ask before you allow it.

Start from the questions you ask each week

Write down the questions you and your team already ask each week, and who decides on each answer. The figures to show follow from those questions.

  • What needs our attention this week? The figures that need a decision are ready before the meeting.
  • Do we have enough stock to deliver? Demand and available stock read together in one view.
  • What is driving the change in sales? From the total down to the customers, regions and products behind it.
  • Who still needs to approve this spend? The request, its reason and its approval history sit in one place.

These four questions come from one of our client stories, where the information sat in several systems. Before, the team pulled separate reports, updated a workbook and checked the numbers before planning could begin. Now demand, stock and constraints are read together in one view, and the planner keeps control of the quantities.

When a business already has Claude, Copilot or Gemini, its first dashboard can be built in that tool. Some questions need figures from several systems at once, or a view the whole team opens each morning. Then we build the dashboard as its own tool, and it reads the systems where the data lives. Our page on how AI can prepare the decisions waiting on you shows the step after it.

How we start

We set up Claude for the businesses we work with, or the AI tool a team already uses, with the connections each person needs and access that matches each role. Our guide on adopting AI securely shows how.

For a dashboard, we start from the questions in the meeting and the systems that hold the answers. We check each figure against those systems before anyone relies on it.

How we help, who we are and how we handle security. On the blog we cover Claude Data and Claude Sheets and Whiteboards, and how to make business information accessible to your team. If you are working out where AI fits first, see where to start with AI in 2027.

The first step with us is a free AI pre-audit

45 minutes to explore where AI could help your business and what to put in place first. With or without us.

Book your free AI pre-audit

Sources

Every fact about a product on this page comes from the vendors’ own announcements and help pages, read on 22 November 2026. The screens are the vendors’ own demo material. Features change often, some of them are in beta or in preview, and details may change after publication.

  1. Anthropic: Build live dashboards and animate explainers with Claude, 8 October 2026
  2. Claude help centre: Get started with Claude Data
  3. Microsoft Learn: Copilot for Power BI overview
  4. Google Docs Editors Help: Collaborate with Gemini in Google Sheets
  5. Google Workspace Updates: Generate charts and valuable insights using Gemini in Google Sheets, 29 January 2025
  6. Google Workspace: AI tools for business
  7. Google: Introducing Gemini 4 Hydrogen, 12 November 2026

Bring your questions.

Free AI pre-audit