Fabric Weekly

Issue 12

29th September 2026

🎉 Welcome to issue #12 of Fabric Weekly!

It's FabCon Europe week in Barcelona, so this edition covers what's been announced so far...

This week, on the official side:

Outside of FabCon...

From the community:

And, some events in the next couple of weeks:

  • Wednesday 30 September, Barcelona - at FabCon Europe? PBIMCR are hosting networking drinks from 6pm. Other UK user groups are also in attendance, so keep an eye out!
  • Wednesday 7 October, Birmingham - #PBIBRUM with Victor Rivas and Jerome Wittersheim.
  • Tuesday 13 October, Manchester - PBIMCR at Data Decoded with Rob Sewell and Duncan Boyne (free Data Decoded ticket needed).
  • Tuesday 13 October, online - Devon & Cornwall Power BI with Cristi Prifti on MCPs, Skills and Plugins.

🤖 Copilot, AI & Agents

  • FabCon and SQLCon 2026 in Barcelona: Building the data foundation for Microsoft Copilot and agents by Arun Ulag · 2 min read Arun Ulag's overview of the FabCon and SQLCon Barcelona announcements: Fabric IQ now generally available in Microsoft Copilot Chat and Cowork, agentic app creation coming to Power BI Desktop (with Fabric Apps and up to 1 GB of Fabric Database per app included for Power BI Pro and PPU customers), IQ sharing and new OneLake integrations, the Osmos-based data engineering agent, observability and on-demand billing in Fabric, and Database Hub, database agents and new Azure SQL Hyperscale capabilities.
  • Bringing governed analytics into the flow of work: Fabric Analytics at FabCon Europe 2026 by Bogdan Crivat · 9 min read Bogdan Crivat's Fabric Analytics round-up: Fabric IQ is now generally available in Microsoft Copilot Chat and Cowork, grounded in existing semantic models, with Fabric IQ MCP GA and data agents GA in Copilot Studio. It also previews app building in Power BI Desktop (Fabric Apps to be included with Power BI Pro), semantic views in OneLake, GPU-accelerated query execution in Fabric Data Warehouse, the Osmos-based data engineering agent, Spark Runtime 2.0 (Spark 4.0, Delta 4.0), GA for Materialized Lake Views, and sensitivity labels and DLP flowing into Copilot with a Restrict from Copilot control.
  • Trusted AI starts with Microsoft Fabric, Real-Time Intelligence, and IQ by Yitzhak Kesselman · 10 min read Yitzhak Kesselman positions Fabric as the live operational layer of Microsoft IQ and lists the Real-Time Intelligence and Fabric IQ updates: Observability Insights and the operations agent's Investigator (both preview), custom streaming connectors and a mirrored database change feed connector for Eventstream, Copilot data exploration and Fabric Maps in Real-Time Dashboards, an Eventhouse MCP server, Business Events, and a long list of ontology additions including AI generation, semantic models as a source, RDF import and export, versioning, MCP tools and OneLake security enforcement.
  • The agent tests its own work by Aleksi Partanen The video demonstrates an AI agent autonomously building a solution, verifying its work, and conducting a full end-to-end test in the bronze layer of Microsoft Fabric using Claude Code.
  • Introducing Fabric-RLM by Sandeep Pawar · 9 min read Sandeep Pawar introduces Fabric-RLM, an open-source Python library built on DSPy's Recursive Language Model module that writes and runs Python inside a Fabric notebook to complete plain-English tasks over Lakehouse files, Delta tables and semantic models of any size, returning only results that pass typed output contracts and validators. In his example, gpt-5-mini via Fabric-RLM built a correct, formatted Excel report from a 1.5M-row IMF CSV that gpt-5.1 couldn't answer directly.
  • Fabric RLM: Why Harness Matters by Sandeep Pawar · 2 min read Sandeep Pawar shows why a harness matters: asked how many days of the week contain the letter 'd', eight of nine small and mid-size models answered wrongly on their own, but all nine got it right through Fabric-RLM, which lets the model write and run Python rather than guess from tokens.
  • Fabric's Data Agent Ran on a GPT Starting With 3 (Allegedly) by Aleksi Partanen A clip from the August Fabric feature summary noting that the Fabric data agent's orchestrator has been upgraded to GPT-5.1, plus Teemu Multanen's unverified rumour that the previous version was a GPT-3-series model.
  • Microsoft Tools for Making Data AI-Ready by James Serra · 7 min read In part two of his AI-ready data series, James Serra maps Microsoft tools to the problems they solve: Dataflow Gen2, notebooks and pipelines for cleaning data; semantic models and Purview for business meaning and governance; Fabric IQ Ontology for shared concepts; Prep data for AI and Data Agent configuration for AI-specific context; and Azure AI Search and Foundry IQ for document retrieval. His advice: work backwards from a specific failure rather than adopting every tool.
  • My Two AI Agents Are Not Allowed To Talk To Each Other by Aleksi Partanen The video shows a secure workflow using two isolated AI agents on Windows and Linux communicating only via Azure DevOps pull requests to avoid contamination in agentic CI/CD processes.

📊 Reporting & Insights

  • From prompt to production: What's new in Fabric Apps by Sachin Patney · 7 min read Fabric Apps, now used by more than 15,000 developers, gains direct connectors to Fabric SQL databases, warehouses, lakehouse SQL endpoints and semantic models (running as the signed-in user, so row-level security applies), PostgreSQL alongside SQL Server, code-defined storage, TypeScript backend functions with a Secret Store, built-in app metrics, a Universal App template and Rayfin plugin for coding agents, and private-by-default access controls.
  • Planning in Microsoft Fabric: From insight to action by Dipti Borkar · 7 min read Planning in Fabric IQ adds direct semantic model connections using the planner's own identity, easier writeback of plans into the semantic model and ontology so Copilot and agents can reason over intent as well as actuals, a Native Planning Engine in preview (Rust and Apache Arrow, up to 10x scale), event triggers that kick off pipelines, dataflows or notebooks on writeback, planning sheets embedded in org apps, multi-driver scenario modelling, and new admin controls over planning sessions.
  • Improvements To Power Query Integration In Power BI Report Builder by Chris Webb · 1 min read Chris Webb highlights under-the-radar Power Query improvements in the latest Power BI Report Builder: the Get Data button is now labelled Power Query, and a new dataset property tab shows the M code, M parameters and the Cloud Connection ID, with a shortcut to Manage Connections and Gateways in the Fabric portal.
  • Microsoft Fabric Apps: Business vs Data Apps – What’s the Difference? by Soheil Bakhshi · 1 min read Soheil Bakhshi compares the two Fabric Apps patterns: operational business apps backed by a Fabric-managed SQL database and GraphQL, and data apps that query an existing Power BI semantic model through the Execute Queries REST API. He covers the tenant setting data apps need, sharing and semantic model permissions (consumers now need only Read, not Build), licensing differences, API throttling and anonymous access — and when a standard Power BI report is still the better choice.
  • Power BI: Pseudo-Clear Slicers When Selections Return No Data by Dane Christian Belarmino Dane Christian Belarmino shows how to keep a Power BI category slicer displaying the relevant category when a combination of selections across other slicers returns no rows, so slicers stay usable instead of appearing to lose their selection. PBIX included.
  • Start using DAX with AI by Marco Russo · 1 min read Marco Russo argues that in 2026 using AI to write DAX has become genuinely effective, and announces three new SQLBI video courses — a free AI for Power BI: Intro, plus DAX with AI: Essentials and DAX with AI: Scenarios — alongside a longer post on what has changed since 2023.
  • Model Your Data in Power BI: Star Schemas, Filter Flow & Field Lists by Reid Havens · 1 min read Reid Havens covers Power BI modelling fundamentals: why a model beats one flat table, how filters flow from dimensions to facts, why star schemas beat snowflakes on query speed, DirectQuery latency, DAX complexity and discoverability, how every visual's query boils down to filter, group and summarise, and how to tidy the fields list with friendly names, display folders, hidden keys and descriptions.
  • 6 Power BI Matrix Features You Probably Missed by Chandeep Chhabra · 1 min read Power BI added six valuable features to Table and Matrix visuals including Visual Calculations, custom totals, better column width control, Auto Expand, frozen row headers, and sentence-based tooltips improving reporting clarity and dashboard usability.
  • The hidden complexity of a daily average in DAX by Alberto Ferrari · 1 min read Alberto Ferrari shows how a seemingly simple daily average in DAX can return different values depending on how edge cases are handled, and uses AI to speed up building the intended calculation.
  • PBIP, PBIR, and Making Power BI AI-Addressable: A Conversation with Rui Romano (Fabric Insider Ep. 14) by Reza Rad · 13 min read Reza Rad talks to Rui Romano about PBIP, PBIR and TMDL: why PBIX isn't going away, why PBIR is becoming the default report format and is the last gate before PBIP reaches GA, and how that same hardening work now underpins agentic Power BI development — the Desktop IPC bridge, the modelling MCP evolving into an authoring MCP, and customisable skills.
  • Skills for Report Authoring in Power BI: A Conversation with Harleen Kaur (Fabric Insider Ep. 15) by Reza Rad · 10 min read Reza Rad talks to Harleen Kaur (Power BI Core AI team) about the new Power BI report authoring skills — planner, design and authoring — that take an AI agent from a semantic model to a themed, multi-page report, plus Desktop Bridge, which lets the agent open, screenshot and self-correct the report in Power BI Desktop. Both are in preview.

⚙️ Data Engineering

  • Fabric September 2026 Feature Summary by Katie Murray · 122 min read The September 2026 Fabric Feature Summary showcases numerous enhancements including governance, CI/CD upgrades, AI-driven experiences, and advanced data integration aimed at building scalable data estates.
  • FabCon/SQLCon Barcelona 2026: What’s new in Fabric Data Factory by Faisal Mohamood · 11 min read Data Factory announces more than 45 new capabilities: Mirroring for Google BigQuery and SharePoint Lists (GA), extended Mirroring capabilities including Change Data Feeds (GA) and Snowflake security role replication, Copy job writing to Eventstreams and triggered by Activator, Dataflow Gen2 Shared Queries and new destinations including Excel and email, Power Query Online experiences coming to Power BI Desktop, dbt Job GA, Lakehouse Maintenance and Approval pipeline activities with exponential retry, and a guided upgrade path from Azure Data Factory.
  • Modernize your Azure Data Estate with Microsoft Fabric: A guided path from Azure Data Factory and Azure Synapse by Faisal Mohamood, Bogdan Crivat · 7 min read A guided route from Azure Data Factory and Synapse to Fabric: ADF customers can now bring an existing factory into Fabric to view and monitor its pipelines, run a compatibility assessment, and upgrade supported pipelines incrementally, with optional partner help. Synapse Spark and dedicated SQL pool migration assistants cover notebooks, Spark pools and warehouses, now with selective and incremental warehouse migration, and the Osmos data engineering agent (preview) can take on the refactoring work.
  • Connecting apps, databases, and AI on one foundation: new innovations at SQLCon and FabCon EU 2026 by Shireesh Thota · 6 min read Shireesh Thota's databases round-up: new Fabric Apps capabilities (data connectors, PostgreSQL, backend functions), Database Hub in preview as one control plane for SQL Server, Azure SQL, PostgreSQL, Cosmos DB and SQL database in Fabric, and database agents for SQL and PostgreSQL (preview). Beyond Fabric, SQL Server on Azure Local is GA, and Azure SQL Hyperscale gains serverless auto-pause and GA DiskANN vector indexing.
  • Extracting Lineage Information from Fabric Spark Notebooks by Gerhard Brueckl · 3 min read Gerhard Brueckl extracts table-level lineage from Fabric Spark notebooks by parsing the leaf nodes of the Spark execution plan (which resolves nested views), maps schema-enabled lakehouse IDs back to readable names, and stores source and target dependencies to drive notebookutils runMultiple() orchestration automatically. Code is on GitHub.
  • A new migration tool every single time by Aleksi Partanen In a clip from the August Microsoft Fabric feature summary, Oskari Heikkinen reflects on how every SQL Server migration — first to Azure, now to Fabric — seems to arrive with a brand-new migration assistant to learn.
  • Does anyone actually tune Spark driver and executor memory? by Aleksi Partanen Spark users rarely adjust driver and executor memory, finding better results in fixing code or scaling clusters.
  • The Free Fabric Speed Toggle You're Probably Not Using by Aleksi Partanen Oskari Heikkinen reports that enabling the Native Execution Engine in Fabric Spark cut one of his pipelines from nine minutes to seven with no code changes and no extra cost, in a clip from the August Fabric feature summary.
  • Your Source System Deletes Data Every Single Day by Aleksi Partanen Aleksi Partanen explains why a source system that overwrites values in place is effectively deleting history every day, and how building SCD2-style change history (valid from, valid to, is-current) into the bronze layer of a Fabric medallion architecture keeps every version for historical analysis.

🏛️ Storage & Platform

  • FabCon and SQLCon Barcelona 2026: What’s new in Microsoft OneLake and its rapidly growing ecosystem by Dipti Borkar · 16 min read OneLake announcements from Barcelona: IQ sharing (preview soon) for sharing governed data and context such as agent instructions and ontologies; Salesforce Data 360 and lakeFS shortcuts; Business Central and BigQuery mirroring; Azure Databricks native storage on OneLake in production; on-demand billing for OneLake compute and the F0 entry point; OneLake catalog in Excel and Microsoft Foundry; the OneLake Table Read API (preview); a ClickHouse workload; Snowflake security role replication (preview); and a batch of partner workloads.

🔐 Governance & Security

  • On-premises data gateway September 2026 release by Goutham Upadhyaya · 2 min read The September 2026 on-premises data gateway (3000.334) adds soft delete, keeping deleted gateways recoverable for 30 days, upgrades the embedded Log4j library to 2.26.1 and addresses CVE-2026-18401, fixes a sign-in issue on Windows Server 2016, and aligns with the September Power BI Desktop release.

💰 Management & Cost

  • Build, deploy, and govern Microsoft Fabric at scale by Kim Manis · 11 min read Platform announcements for developers and admins: observability in Fabric with unified workspace monitoring, a refreshed Monitor Hub and agentic pipeline investigation (preview); deployment plans (preview); bulk item definition APIs (GA) and Git enhancements; on-demand billing and a new zero-provisioned F0 entry point (coming weeks); capacity overage and F4096/F8192 SKUs (GA) with workspace surge protection GA soon; attribute-based policies in Fabric (preview); and outbound access protection extended to OneLake shortcuts (GA).
  • #5 Adding users, licenses and guest access by Paul Turley · 1 min read Part five of Paul Turley's Getting Started & Tenant Foundations series covers giving people access to Fabric: licenses are assigned in the Microsoft 365 admin center (individually, in bulk, or automatically via Entra group-based licensing), external users are invited as Entra B2B guests rather than created, guests on Pro or PPU workspaces still need a license from one side or the other, and guest access only works when both the Fabric admin setting and Entra's external collaboration settings allow it.
  • Storing Fabric Capacity Operation Events into a Lakehouse by Gilbert Quevauvilliers · 1 min read Gilbert Quevauvilliers uses a Python notebook to query Eventhouse and store Fabric capacity operation events in a Lakehouse, an approach chosen because it runs quickly and consumes very few capacity units.
  • Fabric CLI Explained: Install, Use, and Connect It to GitHub Copilot by Reza Rad · 7 min read Reza Rad shows how to install Fabric CLI (including a PATH fix), log in and use commands such as fab ls, fab export and fab cp on their own, then connects it to GitHub Copilot in VS Code so the agent can export a workspace, create a new one and copy content between workspaces — writing its own loop when the simple copy fails.
  • Native vs Custom Branching Out in Microsoft Fabric by Kevin Chant · 1 min read Kevin Chant compares Microsoft Fabric's native branch-out-to-workspace options with Microsoft's custom Branch Out to New Workspace scripts, prompted by the new branch workspace admin configuration settings.
  • Why So Many Microsoft Fabric Projects Fail by Aleksi Partanen The video explains common pitfalls-such as lacking governance, improper workspace naming, and insufficient CI/CD practices-that cause many Microsoft Fabric projects to fail despite its easy setup.

📅 Community & Events

Get Fabric Weekly in your inbox every Tuesday

Sign up to receive the email every week and keep on top of all the announcements.

Don't forget to confirm your subscription. We have a double opt-in sign up process to prevent spam.

We will never provide your details to any third parties. We don't spam.

Fabric Weekly Logo