This month, I'm going to go back to a single blog with all features. Unfortunately trying to produce separate blogs each month is proving too much work. Grab a cup of tea, and get comfortable, with everything from Ignite this will be a long one.

First up, Power BI.

Power BI

Copilot and AI

Ask anything with standalone Copilot updates and availability in Power BI mobile (Preview)

This is the new Copilot interface from the service brought into the mobile app. I still haven't had chance to get hands on with this experience, but it is the single biggest shift that I've seen in the Power BI side for years.

I know many a c-suite that will be excited by this, especially now it can be used on the move between meetings.

Improvements to verified answers

It might seem a small one, but moving from the filter context at time of creation to filter context at the time of consumption is a big step forwards. Add to that Copilot getting better at matching user prompts to verified answers under the hood and this is a significant update to improve reliability of responses.

MCP server for chat with your data (Preview)

Not quite available yet, but previewed at Fabcon in Europe is the MCP server for Power BI. It'll let you get the schema of a semantic model, generate the query and execute it.

The full MCP server that was shown is much more powerful than this, best thing to do is watch this space.

Reporting

Automatically expand matrix columns

This is a really minor change, but one that improves the UX. Having the white space on the visual when you only have a few columns used to be frustrating. This is an improvement.

Card visual (GA)

A long overdue overhaul of the card visual that improves report UI. Well worth updating next time you have to touch a report.

OneLake catalogue supportinng Translytical task flows (Previews)

Makes it easier to find the user functions needed to build capabilities for Power BI to writeback to databases stored in Fabric.

App Copilot (preview)

Copilot is now available in Power BI apps. Given that this is typically how reports are distributed across most organisations, this is going to make it easier to find this information.

Modelling

Semantic model version history (Generally Available)

For those familiar with Notebooks in the Fabric service, this is the Power BI equivalent of that version history functionality. Today it covers semantic models both in the desktop and service.

TMDL Visual studio code extension (GA)

A Microsoft approved plugin for Visual Studio is now avaible. It makes sure that Visual Studio correctly highlights and formats TMDL to make it easy to read and navigate.

Fabric

Fabric platform

Govern in OneLake for tenant admins (Preview)

This is the unification of all the existing reports with the addition of Copilot support. The idea is to make it easier for admins to govern their tenants by simplifying workflows.

Right-click menu for easier multitasking

This one's a simple UX change. I haven't been able to test it so far, as for some reason my tenant still doesn't have the new multitasking experience.

Aure DevOps (ADO) service principle and cross tenant support (GA)

At last we get service principle support for ADO, no need to setup dedicated user accounts in Entra. For those that have multi-tenant setups, you can now have one ADO setup that all tenants can access instead of having to have ADO configured in each tenant - reducing the management overhead this forced.

OneLake

OneLake diagnostics (GA)

Records who accesses what, when and how across Fabric workspaces into a Lakehouse to make it easier to record. This isn't free and will consume CU. So do be aware as if you have a lot of users, it could get expensive very quickly.

To be safe, it is probably worth looking at a diagnostics capacity that powers your central workspace. That way it isn't going to impact your backoffice processes nor end users.

OneLake security ReadWrite permission (Preview)

This one shows just how early in preview OneLake's new security model is. Giving it ReadWrite access controls in a lakehouse should be part of MVP for production.

Keep watching, but stay away from it for now.

Databases

Fabric (SQL) databases (Generally Available)

This is a major milestone for the platform. Given it has been announced at the same time as SQL server 2025 pretty much confirms my suspesion that it was built around that platform.

The big thing to rember that Microsoft haven't increased the SLA on Fabric (it's still 3 9's in the documents). If you require availability beyond this, you need to look into how the high availability and BCDR options impact your architecture.

SQL Auditing (Preview)

Whilst we can now deploy Fabric databases into production, be aware that auditing isn't. That means if you have specific auditing requiests you are going to need to hand-roll an auditing solution still.

Customisable Point-in-time backup retention (GA)

By default Fabric databases are backed up for 7 days. If this doesn't work for your setup, this can now be changed to between 1 and 31 days inclusively.

Customer-managed keys in SQL DB (Preview)

If you need custom encryption keys in your architecture, you can do it but it is a preview feature.

New tools in Copilot sidecar chat (GA)

We've got a bunch of new functionality in Copilot for Fabric DBs - such as being able to find out resource intensive queries. The blog post doesn't give a full list, meaning you'll need to check out the associated learn pages.

Data Virtualisation support for Fabric databases (Preview)

This one doesn't sound key, but for me it is. By default, Fabric DB loads data into its engine and behind the scenes replicates it into OneLake for access by other Fabric items such as Lakehouses.

But what it can't do is read data from OneLake without loading data into the Fabric DB. That's no longer the case with this feature. Fabric DB now supports accessing parquer, delta and csv data using either OPENROWSET or External Tables.

Microsoft Python Driver for SQL database (GA)

For those writing Python scripts against Fabric DB's, you now have a native connector that should be more performant.

Cosmos DB (GA)

Again another major milestone having this available within the platform. As with Fabric DBs, do be aware that the standard SLA is 3 9's and you will need to architect accordingly if this is insuficient.

Data engineering

Spark connector for SQL databases (Preview)

This change simply means that you can access SQL DB (including Fabric DB) from within Fabric notebooks.

ArcGIS GeoAnalytics for Spark (GA)

This is a feature I've been watching for a while. With ArcGIS being the gold standard in spatial analytics, it's great to have native compatibility.

Back in the days of Power BI combining spatial insight with 3rd party data available has produced new insights that wouldn't have been obvious otherwise.

Definitely a feature that I'll use when it is appropriate to map data.

Faster notebook loading with progressive rendering

This one means that notebooks will open quicker.

Optimal refresh for materialised lake views (Preview)

Personally I can't wait to see materialised lake views go GA. It's a much needed feature.

This feature is designed to improve refresh performance by programmatically identify the refresh approach. Switching this feature on let's the engine control if full, incremental or no refresh approach is taken.

VS code improvements

Couple of new features added. If you use Vs code do go check out the changes.

New features in Fabric User data functions

This covers multiple features across the platform:

Fabric activator support (preview). UDFs can now be called when an Activator rule is triggered.
Variable Library integration. They can now be called from UDFs.
Azure key vault support. Key vaults can now be called using UDFs.
Cosmos DB support. By connecting to the endpoint.

Azure artifact feed in Fabric environment (preview)

For that use the Azure artifact feed to install Python libraries, this is now supported in Fabric.

As always, ensure you are using the Spark aware version of Python libraries out of preference. The reason is that using standard Spark libraries will cause the code to only execute on the cluster controler, rather than the entire cluster you have available.

Data science

Connect Data Agents to your Azure Search index in Microsoft Foundry

For those with unstructured data within OneLake, this is the preferred approach to be able to search unstructured data and combine it with structured data to provide a unified view.

Check out the configure your agent documentation on Microsoft Learn to find out how.

Updated Data Agent Limits

The token limits on example queries and Event KQL databases have been increased 3x.

AI functions (GA)

AI functions within Fabric have gone GA. This means that doing complex functions such as clasify, summmarise,etc can now be done as python functions.

MCP support in Fabric data agents

For those that aren't familiar with MCP servers, have a read of my previous article. The support for this protocol will make it easier to integrate Fabric data sets as a tool available for LLM based solutions.

Fabric data agents integrated with M365 copilot

This appears as a option to enable when publishing your data agent within Fabric. Once enabled it can be accessed anywhere that the M365 copilot can be accessed.

Prep your data for AI for semantic models

Fabric data agent now fully supports prep for AI customisations for semantic models. Making it easier to guide your AI onto which tables to use, terminology, etc. Given that data agents are still in preview, I assume this functionality is too.

Fabric data agent database support

Fabric/SQL databases are now added as direct support for data agents and we've had some UX/UI changes.

ML tracking upgrade

It sounds like a bit of a major upgrade to MLOps in Fabric is coming. If you are using the ML tracking make sure that you schedule in the upgrade to take advantage of the upcoming changes.

Internal Python Packages in ML Model Endpoints

Previously AutoML and FLAML packages used internal packages and so couldn't be deployed to an ML model endpoint. That's now been fixed

Data Warehouse

IDENTITY() columns (Preview)

Great that this is now in preview, should really been part of GA, but at least it is fixed. Just hope it doesn't sit in preview for the next year ...

Data clustering (Preview)

For those from an onprem background, this is basically the equivalent of clustered column store indicies. Great to see it supported, but again another feature that is a must have for most large scale enterprises.

Warehouse snapshots (GA)

This one is actually a pretty big feature. Again, harking back, I remember the days when you'd have to go dig out an old backup and restore it to a database with a different name.

Now that doesn't need to happen. Instead, Snapshots can be taken to provide a read only view - basically the equivalent of delta's timetravel but across the entire warehouse.

For debugging issues with ML models, reports, etc, this one is pretty big.

Varchar(MAX) support

Another feature that should have been GA, not much more needs to be said on that.

RTI

New connectors

This month we get CDC for Mongo DB, Cribil, and HTTP connector. It's the last of these two that is worth paying attention to.

The new HTTP connector effectively allows data to be streamed directly from any Rest API. I need to get hands on with this, but it could mean that building custom API integrations in order to get data out of 3rd party platforms becomes a thing of the past.

It sounds too good to be true, so need to understand the limitations. For example, I don''t know how it deals with paginated APIs.

The main issue though is the lack of git integration and deployment pipeline support. Till that's sorted, I would stay well away from it for production purposes.

Data activator as destination (GA)

For those that need action taken directly off the back of events hitting an eventstream, you now have the capability to route straight into data activator.

Capacity Overview events (preview)

For those wanting to build custom capacity reports, you can now use RTI to canputre a capacity sunnary and the current capacity state.

Entity diagram in Eventhouse (Preview)

Easiest way to think of this is lineage view for RTI, but with more detail. It'll track from the eventstream that lands the data into the Eventhouse through to continous exports - whilst covering relationships, functions, materialized views, etc in between.

Operations agent (Preview)

This is one of the big ticket announcements from ignite. These agents sit over an Eventhouse and use instructions and business goals to work out events to watch for.

If you want to test it out, this one is paid capacities only unfortunately.

Imagery file support and data labeling (Preview)

For those that heavily use specific spatial formats, these can now be uploaded and natively displayed within a lakehouse. On top of that, critical attributes can be labled natively.

Copilot analysis support in RTI (Preview)

Copilot has now gained RTI dashboard support.

Data Factory

Enterprise readiness

This group covers a set of features that are specific to certain architectures. If they impact you do go check out the release notes:

Key pair authentication for Snowflake connector
One click update for On-premise gateway - instead of having to run an upgrade file on the server this can now be updated from the service UI. You must be running the November 2025 version of the gateway for this to be available
VNET Data Gateway gains certificate and proxy support (Preview)
Connection paramaterisation with variable library for CI/CD (GA)

Copilot improvements

This month sees a few enhancements to Copilot for pipleines. These cover:

Generating summaries of errors for failed pipelines
Improvement to expression builder for Copilot
Ability to get Copilot to explain complicated expressions

UX improvements to Monitoring hub

A hierarchical view has been added to make it easier to see upstream and downstream jobs in a pipeline, as well as identify dependencies.

Beyond this pipelines have seen a number of connector updates for spark and impala 2.0, and various changes to the copy job.

Dataflow changes

For those using Dataflows, this month we have seen:

UX changes to get data in Excel
UX changes to getdata from Copilot
AI function integration
Power Query Intellisense support

Apache Airflow

For those using Apache Airflow's item available within Fabric, we've seen a few new items this month:

New Job API to manage files and requirements
Ability to upload Airflow project files via the UI
New job management APIs to manage DAG files

Mirroring

Mirroring has seen the following updates this month:

Iceberg table support
SAP Datasphere (preview)
SQL server 2016-2025 (GA)
Cosmos DB (GA)
Azure SQL DB user assigned managed identity support (Preview)

November 2025 - Fabric and Power BI update blog