Have you ever ever wished to construct an interactive dashboard immediately on prime of your current Energy BI semantic fashions, with out opening Energy BI Desktop or writing a single line of DAX?

With the introduction of Claude Dashboards, that is now attainable. We are able to join Claude to a Energy BI semantic mannequin, describe the dashboard we would like in plain English, and let Claude generate the visualisations and the DAX queries behind them.

Sounds fairly good, doesn’t it?

I not too long ago spent a while experimenting with Claude Dashboards, connecting them to Energy BI semantic fashions by way of Microsoft’s Energy BI Mannequin Context Protocol (MCP) server. The outcomes had been fairly fascinating. I managed to construct interactive dashboards utilizing the measures, relationships and enterprise logic already outlined in my semantic fashions.

However getting a dashboard to work is one factor. Making it sensible for an actual enterprise atmosphere is a distinct story.

What occurs after we share the dashboard with different customers? Does Row-Stage Safety (RLS) nonetheless work? What permissions do customers want on the semantic mannequin? And might we join Claude to a number of Microsoft Cloth tenants? These are a few of the questions I wished to reply.

On this weblog, I am going by way of the setup step-by-step, clarify how the connection works, and share what I discovered alongside the best way. There are a number of gotchas round authentication, safety, sharing and connecting to a number of tenants that we have to perceive. I examined this in October 2026, utilizing the beta model of Claude Dashboards.

You’ll be able to watch the whole step-by-step demonstration on my YouTube channel:

In case you desire listening, I’ve additionally printed an AI generated podcast model of the video:

What Are Claude Dashboards? #

Claude Dashboards is a function inside Claude’s Artifacts expertise. We are able to describe the dashboard we would like in plain English and join it to exterior knowledge platforms and purposes.

The expertise jogs my memory of constructing an internet utility with AI. We describe what we would like, Claude builds the dashboard, and we will hold asking for modifications in the identical dialog.

The fascinating half is that we’re not restricted to importing Excel information or CSV extracts. We are able to connect with exterior knowledge platforms, together with Energy BI semantic fashions by way of a customized connector.

Claude generates DAX queries towards an current semantic mannequin and makes use of the outcomes to show the information within the dashboard. We’re not creating a brand new semantic mannequin, rebuilding relationships, or manually exporting the whole dataset.

Having mentioned that, there is a crucial level right here. Not exporting your entire dataset doesn’t imply that no knowledge leaves Energy BI. The DAX question outcomes go by way of the MCP connection for Claude to course of and show. We have to hold this in thoughts when working with delicate enterprise knowledge.

Is a Claude Dashboard interactive? #

One of many questions I seen on social media shortly after the discharge of Claude Dashboards was whether or not the generated dashboards are interactive. The reply is sure, they are often.

Nonetheless, what Claude builds is an internet utility. We should always not count on it to robotically embrace all of the interactions we’re used to in Energy BI visuals.

If we would like an interactive dashboard, we should always explicitly ask for it. For instance, we will request dropdown filters, interactive charts, and visualisations that reply to choices on different components of the web page.

This labored in my check, though a number of of the visuals didn’t behave fairly how I anticipated.

How Claude Connects to Energy BI Semantic Fashions #

Watch the structure clarification on YouTube at 4:45

As of my October 2026 testing, Claude didn’t present a devoted built-in connector for Energy BI semantic fashions. As a substitute, I used a customized connector pointing to Microsoft’s hosted Energy BI Consumption MCP server.

The next diagram reveals what occurs, from organising the connector to displaying the dashboard:

sequenceDiagram
    autonumber
    actor Person as Dashboard creator
    participant Claude as Claude Dashboard
    participant Connector as Customized connector
    participant Entra as Microsoft Entra ID
    participant MCP as Energy BI MCP server
    participant Mannequin as Semantic mannequin

    rect rgba(120, 120, 120, 0.12)
    Notice over Person,Entra: One-time setup: join Claude to Energy BI
    Person->>Connector: Add the MCP server URL and the app consumer ID
    Connector->>Entra: Delegated OAuth sign-in
    Entra-->>Connector: Entry token for the signed-in account
    Notice over Connector,Entra: This account turns into the question id
    finish

    rect rgba(120, 120, 120, 0.12)
    Notice over Person,Mannequin: Constructing the dashboard
    Person->>Claude: Immediate: construct an interactive dashboard of gross sales by area and month
    Claude->>Connector: Learn the mannequin schema
    Connector->>MCP: Name the schema software with the token
    MCP->>Mannequin: Learn tables, measures and relationships
    Mannequin-->>Claude: Schema, returned by way of the MCP server and connector
    Claude->>Claude: Write the DAX queries and the dashboard web page
    finish

    rect rgba(120, 120, 120, 0.12)
    Notice over Person,Mannequin: Each time the dashboard is opened, refreshed or filtered
    Claude->>Connector: ExecuteQuery with a DAX question
    Connector->>MCP: Ahead the question with the identical token
    MCP->>Mannequin: Run the DAX because the connector account
    Mannequin->>Mannequin: Apply RLS for the connector account
    Mannequin-->>MCP: Solely the rows that account may even see
    MCP-->>Claude: Question outcomes depart Energy BI
    Claude-->>Person: Render the charts and filters
    finish

    Notice over Claude,Mannequin: RLS follows the id the connector signed in with. Validate what every shared viewer will get.

The Energy BI semantic mannequin stays chargeable for evaluating the DAX queries, making use of mannequin safety, and returning the outcomes. Claude builds the dashboard interface, prepares or runs the queries, and shows the outcomes.

One half we have to take note of is the account used for the connection. The Energy BI MCP server runs queries utilizing that account’s id. We come again to this when RLS and sharing later on this weblog.

Conditions #

For this experiment, I used a paid Claude plan with entry to Claude Dashboards, and a Energy BI Professional licence with entry to the semantic fashions.

The Claude Dashboards beta is obtainable on Professional, Max, Staff and Enterprise plans. Enterprise directors could must allow the function first.

On the Microsoft aspect, we want an Entra utility registration, permission to configure the required Cloth tenant settings, and entry to the Energy BI semantic mannequin. The Energy BI Consumption MCP server additionally requires Construct permission on the semantic mannequin for executing queries.

The Energy BI licence we want can range relying on the workspace capability and the way customers entry the content material. In my check, I used an account with a Professional licence.

Step 1: Register an Software in Microsoft Entra #

Watch the Entra ID configuration on YouTube at 5:55

Step one is to register an utility in Microsoft Entra ID. Open Microsoft Entra Admin Centre and comply with these steps:

1. Within the left pane, choose App registrations.
2. Click on New registration.
3. As soon as registered, your utility seems within the record. Mine is Claude Dashboard for Energy BI.

On the Register an utility web page:

1. Enter a reputation for the applying. I named mine Claude Dashboard for Energy BI.
2. Below Supported account sorts, choose Single tenant solely. The registration have to be created within the tenant internet hosting the Energy BI atmosphere.
3. Go away Redirect URI empty for now. We add it later beneath Authentication.

Then click on Register.

You want permission to register purposes. Relying in your organisation’s Entra configuration, normal customers could also be allowed to register purposes, or an administrator could must carry out this step.

As soon as registered, copy the Software (consumer) ID from the applying’s Overview web page. We want it when configuring the connector in Claude.

Configure the API permissions #

Subsequent, the applying wants delegated Energy BI permissions. Open the applying registration and comply with these steps:

1. Within the Handle part of the left pane, choose API permissions.
2. Click on Add a permission, choose Energy BI Service, then select Delegated permissions. Discover every permission you want, tick it, and click on Add permissions.
3. The added permissions seem beneath Energy BI Service: Dataset.Learn.All and MLModel.Execute.All.
4. Click on Grant admin consent for Contoso. Contoso is the title of my check tenant; you will note your individual organisation’s title right here.

The quickest strategy to discover a permission in step 2 is the search field. For instance, typing ml finds MLModel.Execute.All:

These are the 2 permissions I ended up with:

There’s a element across the documentation price mentioning right here. The Energy BI Consumption MCP server setup web page lists Construct permission on the semantic mannequin, however doesn’t record Workspace.Learn.All. The separate information for registering the hosted servers with exterior MCP purchasers does record that delegated API permission for each the Consumption and Authoring servers. Each pages seek advice from the identical Consumption endpoint used on this weblog. My examined setup labored with out Workspace.Learn.All, so I eliminated it, however this doesn’t show that each operation works with out it. Additionally, this API permission is completely different from giving a person a workspace function. Through the recording, I seen that I had granted SemanticModel.ReadWrite.All by mistake. The external-client information lists it for the Energy BI Authoring MCP server, so I eliminated it from this consumption-only setup too.

That is one thing I at all times counsel checking when registering an utility. We should always solely give it the permissions it wants for what we’re constructing.

Relying in your tenant’s consent insurance policies, you might also want administrator consent. In my check atmosphere, I granted it in step 4 so the consent course of wouldn’t interrupt the sign-in.

Configure the redirect URI #

Subsequent, we have to configure the authentication callback URL.

1. Within the left pane, choose Authentication.
2. Add a redirect URI for the Cellular and desktop purposes platform, utilizing:

   https://claude.ai/api/mcp/auth_callback

This URL takes us again to Claude after finishing the Microsoft sign-in. Make sure that the redirect URI matches what the MCP consumer expects. If this URL is inaccurate, the sign-in will fail.

For an in depth clarification, see the Microsoft Be taught documentation on registering Energy BI MCP servers with exterior purchasers.

Step 2: Allow the Required Microsoft Cloth Settings #

Watch the Cloth Admin Portal configuration on YouTube at 10:15

Now that the applying has been registered, we have to configure Microsoft Cloth. Open Microsoft Cloth, click on the settings icon, and navigate to the Admin Portal. In my check atmosphere, I enabled two settings.

Energy BI Mannequin Context Protocol Server #

1. Kind mcp within the search field.
2. Develop Customers can use the Energy BI Mannequin Context Protocol server endpoint (preview) and swap it to Enabled.
3. Click on Apply.

This allows customers lined by the tenant setting to hook up with the Energy BI MCP endpoint.

One correction from the video: I initially referred to this because the Cloth MCP server. The endpoint used on this experiment is particularly the Energy BI Consumption MCP server, not the overall Cloth MCP server.

Semantic Mannequin Execute Queries REST API #

The second setting I enabled was:

Semantic Mannequin Execute Queries REST API

That is the Energy BI REST API functionality for executing DAX queries towards semantic fashions.

I had each settings enabled in my check atmosphere. The MCP setup documentation particularly lists the MCP tenant setting as a requirement. So, having each enabled in my check doesn’t inform us that the REST API setting can be required for each MCP configuration.

For testing, I enabled the settings for your entire organisation as a result of I used to be utilizing a devoted check tenant. For a manufacturing tenant, I counsel limiting these settings to a managed safety group the place attainable. In my case, enabling them for everybody was just for testing.

Step 3: Create a Customized Connector in Claude #

Watch the connector configuration on YouTube at 11:45

Now that now we have completed the configuration on the Microsoft aspect, let’s join Claude to Energy BI.

I used Claude Desktop on Home windows, though customized connectors can be found by way of the Claude net expertise too.

Open Settings > Connectors > Add customized connector.

Give the connector the identical title as your Entra utility. On this instance, use:

Claude Dashboard for Energy BI

For the distant MCP server URL, enter:

https://api.cloth.microsoft.com/v1/mcp/powerbi

That is Microsoft’s hosted Energy BI Consumption MCP server endpoint. Don’t confuse it with the Energy BI Authoring MCP server, which makes use of a distinct endpoint and serves a distinct function.

Within the Add customized connector dialog:

1. Enter the connector title.
2. Enter the distant MCP server URL.
3. Click on Proceed.

On the subsequent web page:

1. Below OAuth consumer, choose Use your individual OAuth consumer.
2. Paste the Software (consumer) ID copied from the Entra utility registration. The consumer secret can keep clean.

Then click on Add.

Claude will then information you thru the Microsoft authentication course of. As soon as linked, we will select whether or not Claude can run the instruments robotically or must ask for permission first.

Overview the software permissions and select which instruments can run robotically. I allowed the related instruments to run robotically for this check, however select what is suitable on your atmosphere.

At this level, our connection between Claude and Energy BI is prepared.

Step 4: Construct an Interactive Dashboard Utilizing a Immediate #

Watch the dashboard creation on YouTube at 13:30

Now comes the fascinating half. Let’s construct the dashboard. In Claude, begin a brand new dashboard:

1. Within the left pane, choose Artifacts.
2. Below Make one thing new, choose Dashboard (beta).

For this experiment, I used an current semantic mannequin known as Gross sales & Buyer Insights, positioned within the SalesAndMarketing workspace.

The immediate itself is sort of easy. We inform Claude what dashboard we would like, which semantic mannequin to make use of, the place to search out it, and which connector to make use of.

Enter the next immediate, changing the mannequin, workspace and connector names with your individual:

Construct me an interactive dashboard of gross sales by area and month from the “Gross sales & Buyer Insights” semantic mannequin from the “SalesAndMarketing” workspace utilizing the Claude Dashboard for Energy BI connector

Claude then begins engaged on the dashboard. This took about seven minutes in my check, so the time could also be completely different in your atmosphere. When the dashboard is prepared, overview the request to make use of the connector instruments, together with the DAX question execution software:

1. The This artifact makes use of connectors dialog lists the Claude Dashboard for Energy BI connector and the Execute question software it makes use of. Go away the connector switched on.
2. Click on Proceed.

After approving the connection, verify the dashboard’s gross sales visualisations and filters. Choose a rustic, similar to Germany or the UK, and verify how the information modifications in different components of the dashboard. In my check, these choices up to date the displayed knowledge.

Nonetheless, some interactions didn’t initially behave as I anticipated. One of many area comparability visuals, for instance, didn’t reply in the identical approach as the opposite charts. So, we will construct an interactive dashboard by describing what we would like, however we nonetheless must verify how every interplay works. Getting the dashboard to load is just a part of the work.

Step 5: Understanding Row-Stage Safety #

Watch the RLS investigation on YouTube at 17:00

That is in all probability a very powerful a part of your entire experiment. Once we join Claude to a Energy BI semantic mannequin, we count on the safety guidelines we already need to nonetheless apply. Specifically, if the semantic mannequin has Row-Stage Safety configured, customers ought to solely see the information they’re authorised to entry.

The excellent news is that RLS is enforced by the semantic mannequin when DAX queries are executed utilizing an appropriately restricted person id. Nonetheless, that is just one a part of the safety story.

Why the authenticated id issues #

Once we configure our customized connector, we authenticate utilizing an account from Microsoft Entra. The MCP connection then makes use of that authenticated id to entry the semantic mannequin. So, the RLS guidelines for that account decide which rows Claude receives.

Because of this RLS can work completely for the account utilized by the connector. However what occurs after we share the generated dashboard with different customers? Earlier than answering that query, let’s configure the permissions and check RLS.

Eradicating workspace-level permissions #

For this experiment, I used an account known as MOD Administrator to attach Claude to Energy BI. Initially, this account had administrative entry to the SalesAndMarketing workspace. That is a crucial element as a result of Energy BI doesn’t implement RLS for workspace Admin, Member or Contributor roles.

To check RLS correctly, sign up with one other administrator account and take away the connector account’s workspace entry, together with entry inherited by way of a workspace safety group. I used Alex Weber’s account for this step.

Verify that the connector account now not has a workspace function earlier than persevering with.

Granting permissions immediately on the semantic mannequin #

Subsequent, open the semantic mannequin’s Handle permissions web page and grant the connector account:

• Learn permission
• Construct permission

Go away the choices to switch the semantic mannequin and reshare it unticked.

This provides the account the permissions required to question the mannequin by way of the MCP server with out granting entry to your entire workspace. This additionally reveals one thing helpful: a person doesn’t want a workspace function to question a person semantic mannequin if they’ve the required direct permissions on that mannequin.

Construct permission is required by the Energy BI Consumption MCP server’s Execute Question software. Learn permission alone will not be adequate for this MCP question situation. Having Construct permission doesn’t robotically bypass RLS both. RLS nonetheless applies to customers with Construct permission when they’re topic to its guidelines, so long as they don’t have a workspace function or different permissions that bypass mannequin safety.

Assigning the RLS function #

Subsequent, configure Row-Stage Safety. Open the semantic mannequin, swap to modifying mode, and choose Handle roles on the ribbon.

Assign the connector account to the RLS function you wish to check. On this instance, the function is named Web Gross sales 2026, which restricts the account to knowledge for the 12 months 2026. Within the Handle safety roles dialog, comply with these steps:

1. Swap to the Assign tab.
2. Choose the Web Gross sales 2026 function.
3. Enter the connector account’s e mail deal with and click on Add. The account seems beneath Members.
4. Click on Save.

Shut the dialog and return to Claude. At this level, the anticipated result’s fairly easy. The dashboard ought to now not present gross sales knowledge for earlier years. It ought to solely return knowledge for 2026. And that’s precisely what occurred, however not instantly.

Refreshing the dashboard was not sufficient #

After assigning the function, refresh the dashboard and verify which years can be found. In my check, the outcomes didn’t change at first. From my expertise, permission and safety modifications can take some time to use. However on this check, refreshing the dashboard a number of occasions didn’t change the outcomes.

If the outcomes nonetheless don’t replicate the RLS function, open Claude’s connector settings, disconnect the customized Energy BI connector and join it once more. This was the step that labored in my check.

After reconnecting, refresh the dashboard once more. This time, the outcomes had been restricted to 2026. The 12 months dropdown confirmed solely 2026, which was in step with the newly assigned RLS function. This confirmed that RLS was working for the account I used to attach Claude to Energy BI. You could not must reconnect each time. It was what labored for me on this check.

Earlier than the function was assigned, the dashboard might examine 2026 with 2025. Afterwards, the Yr dropdown solely presents 2026 and the Change on final 12 months card has no 2025 figures to check towards.

The Large Gotcha: RLS, Connector Identification and Sharing #

Thus far, the whole lot sounds fairly good. We now have created an interactive dashboard, linked it to an current semantic mannequin, and verified that RLS can prohibit the information returned by the connector.

However there’s a important distinction between RLS working for the connector’s authenticated account and RLS working independently for each one that opens a shared dashboard.

When utilizing Energy BI stories, we usually count on every person’s efficient id to be evaluated towards the mannequin’s safety guidelines. For instance, if two gross sales managers have completely different regional permissions, every ought to see solely their very own area. An identical expectation exists when sharing a Claude Dashboard backed by a Energy BI semantic mannequin.

What I discovered in my experiment #

The dashboard used the id authenticated by way of my customized Energy BI connector. Once I restricted that id to 2026, the dashboard mirrored that restriction. Nonetheless, I didn’t full a check the place every particular person opening a shared dashboard signed in individually and queried the Energy BI semantic mannequin utilizing their very own account. Nonetheless, it’s extremely possible that the credendials used through the customized connector registration is the one that’s used for refreshing the information for the customers with the shared hyperlink, no less than it appears to be the best way taht the customized connector structure works as we speak.

So, we have to be cautious with the conclusion right here. I confirmed which account my connector used, however this doesn’t show that each shared Claude Dashboard at all times makes use of the creator’s account. Extra importantly, the best way the customized connector behaves as we speak, doesn’t essentially imply it will be the identical structure used when the long run official connector for Microsoft Cloth gadgets is obtainable in Claude.

Actually, Anthropic’s present documentation for sharing artifacts says that artifacts utilizing linked purposes ought to use every viewer’s personal connections somewhat than the creator’s. That’s the supposed behaviour, however it nonetheless must be verified for the actual customized Energy BI MCP connector and dashboard configuration we’re utilizing.

Earlier than utilizing this in manufacturing, I’d check it with no less than two customers who’ve completely different RLS assignments. I’d additionally wish to affirm what occurs when a person has no entry to the underlying semantic mannequin. Till that check works as anticipated, I’d not assume that sharing a working dashboard means every person’s knowledge safety can be working accurately.

The purpose right here is that we have to verify whose account runs the queries and which safety guidelines apply, not simply whether or not the dashboard appears proper.

Step 6: Can We Join Claude to A number of Microsoft Cloth Tenants? #

Watch the multi-tenant check on YouTube at 25:20

One other query I had was whether or not we will create a number of customized connectors in Claude, every pointing to a distinct Microsoft Cloth tenant. This could be fairly helpful for consultants working with completely different buyer environments. In Energy BI and Microsoft Cloth, switching between completely different tenants is sort of frequent. Ideally, we might be capable to create separate Claude connectors for every tenant.

Nonetheless, I got here throughout one other limitation in my check. Once I tried to register a second customized connector utilizing the identical Energy BI MCP endpoint:

https://api.cloth.microsoft.com/v1/mcp/powerbi

Claude confirmed a message saying that the URL already existed within the organisation.

The problem is that the hosted endpoint is similar no matter which Microsoft Cloth tenant we wish to entry. In my examined configuration, Claude wouldn’t permit me to register one other customized connector utilizing the an identical MCP server URL. I additionally couldn’t discover a supported approach so as to add a tenant-specific identifier to the MCP URL so Claude might inform the connectors aside.

Is there a workaround? #

Whereas recording the video, one other concept occurred to me. Microsoft has a number of MCP servers with completely different endpoints, and a few have overlapping capabilities. Maybe we might use a distinct appropriate endpoint for one more tenant.

Nonetheless, I didn’t check this strategy, so I can’t affirm that it really works. We also needs to keep in mind that completely different MCP servers could expose completely different instruments and safety necessities. Utilizing one other endpoint will not be essentially an equal alternative for the Energy BI Consumption MCP server.

So, in my check, I couldn’t register separate connectors for various tenants utilizing the identical Energy BI MCP endpoint in Claude. That’s the limitation I got here throughout with this setup. It doesn’t inform us whether or not one other strategy might assist a number of tenants.

Step 7: Inspecting the DAX Queries Generated by Claude #

Watch the DAX question inspection on YouTube at 27:10

One function I fairly preferred was with the ability to see what occurs behind the visualisations. We are able to open the dashboard’s knowledge sources and take a look at the queries and transformations behind the outcomes. This provides us a greater concept of what Claude has truly constructed.

To examine the queries behind the visualisations, comply with these steps:

1. Click on the information sources button on the prime proper of the dashboard.
2. The Sources panel lists each question behind the dashboard. Click on one to see its particulars.

Right here, you’ll be able to see the DAX queries, their returned outcomes and the transformations utilized to the information. We are able to additionally examine a person visualisation to grasp which question is chargeable for its knowledge. The interface confirmed a preview of the outcomes, restricted to the highest 500 information in my instance.

That is helpful for troubleshooting, particularly when the numbers displayed by a dashboard don’t look proper. Having mentioned that, with the ability to see the DAX question doesn’t imply Claude has written it accurately. We nonetheless must confirm the logic, verify the measures getting used, and ensure the returned numbers are in step with the enterprise definitions within the semantic mannequin.

For me, that is the place a well-designed semantic mannequin turns into much more vital. When our measures, relationships and enterprise logic are properly outlined, instruments similar to Claude have a greater place to begin to work with.

Claude Dashboards vs Energy BI Studies vs Microsoft Cloth Apps #

Watch my comparability and closing ideas on YouTube at 28:00

Now that now we have a working dashboard, it’s time to ask an even bigger query. Are Claude Dashboards going to interchange Energy BI stories? In my view, not of their present kind.

Energy BI stories are far more than a group of charts. They’re a part of a mature analytical platform with established capabilities for report improvement, sharing, safety, governance and enterprise distribution.

Claude Dashboards are fascinating as a result of they permit us to construct analytical experiences in a short time utilizing pure language. However we additionally want to consider how we safe, deploy, preserve and share that dashboard. Producing it doesn’t robotically give us the identical management over these components.

What about Microsoft Cloth Apps? #

This comparability is extra fascinating to me. Once I constructed my first Claude Dashboard, it instantly jogged my memory of Microsoft Cloth Apps. In each instances, we’re creating an application-like expertise round current enterprise knowledge.

In my earlier weblog publish, Microsoft Cloth Apps: Enterprise vs Information Apps, What’s the Distinction?, I defined how a Cloth Information App can use an current Energy BI semantic mannequin with out rebuilding the analytical logic.

Claude Dashboards comply with an identical concept, though how they work and what the platform helps are completely different.

Right here is how I’d examine these choices primarily based on their supposed use and my testing:

One distinction we have to take note of is the permissions on the semantic mannequin. For a Cloth Information App, Microsoft modified the permission requirement for shoppers. Learn permission is now sufficient within the supported sharing situation, whereas the creator nonetheless wants Construct permission. The Energy BI Consumption MCP server utilized by Claude requires Construct permission for its Execute Question software. So, whereas each approaches can use the identical underlying semantic mannequin, we should always not assume they’ve an identical safety and permission necessities.

When Would I Use Claude Dashboards? #

With all these limitations, I can nonetheless see some helpful situations for Claude Dashboards. For instance, an analyst or advisor might use Claude to shortly discover an current semantic mannequin, construct a dashboard for a particular query, or experiment with other ways of visualising enterprise info.

This may be notably fascinating after we have already got a ruled semantic mannequin with well-defined DAX measures. As a substitute of beginning with an empty report canvas, we will describe what we wish to analyse after which verify the outcomes. I can even see myself utilizing it to construct a fast prototype earlier than deciding whether or not the ultimate answer needs to be a Energy BI report or a customized utility.

Nonetheless, if the applying goes to serve many enterprise customers with completely different safety necessities, there may be extra work to do. I’d check authentication, authorisation, sharing, knowledge dealing with and efficiency correctly earlier than selecting this strategy.

The beta appears promising, however getting a prototype working remains to be a fairly completely different factor from having an analytical utility prepared for manufacturing.

Remaining Ideas #

Claude Dashboards are fairly spectacular. On this experiment, I linked Claude to an current Energy BI semantic mannequin, generated an interactive gross sales dashboard utilizing a easy immediate, inspected the DAX queries, and verified that Row-Stage Safety was utilized to the account authenticated by way of the customized connector.

I additionally confirmed that the account didn’t want entry to the workspace. Giving it Learn and Construct permissions immediately on the semantic mannequin was sufficient for this MCP consumption situation.

However I additionally got here throughout some challenges. I needed to register an Entra utility, configure delegated API permissions, change the Cloth tenant settings and create a customized connector. I additionally needed to check the dashboard interactions. The safety modifications solely confirmed within the dashboard after I reconnected the connector. I couldn’t get separate connectors working for a number of tenants, and I nonetheless want to check which account is used for every viewer when sharing.

None of those findings imply Claude Dashboards usually are not helpful. They do present why we have to be cautious about treating a working dashboard as an analytical answer prepared for manufacturing.

Would I take advantage of Claude Dashboards for my very own analytical exploration and prototyping? Completely.

Would I change my current Energy BI reporting atmosphere with them as we speak? No.

And would I think about them a possible competitor to some Microsoft Cloth Apps situations? Sure, notably the place we would like a customized analytical interface constructed on prime of an current semantic mannequin.

However it’s nonetheless early days, and it will likely be fascinating to see how these capabilities evolve.

In case you have been experimenting with Claude Dashboards, particularly with per-user authentication, shared dashboards or a number of Cloth tenants, I’d have an interest to listen to the way it went for you. Have you ever managed to get a distinct configuration working? Please share your findings within the feedback.

Additional Studying #

In case you discovered this convenient, you’ll be able to comply with me on LinkedIn, subscribe to BI Perception on YouTube, or comply with the BI Perception podcast on Spotify. I additionally share updates on Bluesky and X, overlaying Energy BI, Microsoft Cloth, AI, and the real-world knowledge and analytics issues I come throughout alongside the best way.

Associated