Power BI Weekly

Issue 381

30th October 2026

Two official posts to start us off this week. The On-premises data gateway October 2026 release tidies up accessibility in the installer and stops connection tests from resetting your Power Query settings, but the most useful addition for admins (and the one I'm particularly thankful for) is the new Admin Connections REST APIs (in preview), which finally give Fabric administrators visibility of every connection in the tenant rather than just the ones they've been granted access to. Also from the team, Alper Sarikaya has written about Data Visualization in the Agentic Era, introducing new chart and data grid libraries for Fabric Apps that carry over Power BI's visual style and formatting, so agent-built apps don't each reinvent theming and accessibility from scratch.

Over in the community, Marco Russo has introduced version 2 of the DAX LLM Benchmark, built with Maxim Anatsko, which finds that the latest generation of models now writes DAX well enough to save time (wahoo!), with the frontier models performing impressively. On the modeling side, Gilbert Quevauvilliers has shown how to use Direct Lake across public and Private Link workspaces, despite the documentation suggesting this isn't possible. Who doesn't love a workaround?

For the admins (again), Heidi Hasting has written about the fact that the OneLake catalog's Govern tab now surfaces tenant settings in the blog The Govern Tab Shows Tenant Settings Now… Unless You're Not an Admin. Finally, Valerie Junk has untangled the differences between workspace apps, org apps and Fabric Apps, and when you'd reach for each.

Data Prep

  • Build a Fabric IQ Declarative Agent for Microsoft 365 Copilot Chat by Simon Richert · 9 min read Microsoft 365 Copilot Chat can already query Power BI reports and semantic models through Fabric IQ. This post shows how to build a dedicated declarative agent scoped to a curated set of them, skipping discovery lag and stray Office content, adding conversation starters and controlling chart styling. A copy-paste prompt has a coding agent generate the manifest, routing table and Fabric IQ MCP plugin, and user permissions and RLS still apply.
  • Calculating Weighted Average in Power BI by Dinesh Asanka · 3 min read Power BI calculates weighted averages by merging datasets, multiplying marks by weights, summing products, and dividing by total credits, useful in educational assessments with unequal component contributions.
  • Context Engineering Is THE Skill to Learn: Amir Netz | Fabric Insider Ep. 26 by Reza Rad, Amir Netz In Fabric Insider Ep. 26, Reza Rad puts the community's questions to Amir Netz: CI/CD in Fabric, how Fabric F0 works, the role of Ontology alongside the semantic model, using AI well as a Fabric developer, the Osmos acquisition, whether users still need to learn Power BI reporting now Fabric Apps exist, and why Amir calls context engineering THE skill to learn.

Data Modeling

  • Is Power BI Reporting Dead? (What AI Changes for Power BI) by Reza Rad AI transforms Power BI by generating front ends, shifting teams to semantic model ownership, and enabling direct user interaction via AI assistants.
  • DAX LLM Benchmark v2 by Marco Russo · 6 min read Marco Russo and Maxim Anatsko have released version 2 of the DAX LLM Benchmark at daxbench.com, with single-shot tests of measures, queries, calculated tables and columns, and UDFs on Contoso-based models. Six frontier models from three vendors now score between 90% and 96%, and model generation matters more than vendor. Marco argues it is still worth learning to read DAX so you can validate the business rules an LLM implements.
  • Building Claude Dashboards on Power BI Semantic Models: Step-by-Step Setup, RLS, Sharing and Limitations by Soheil Bakhshi · 1 min read Soheil Bakhshi walks through building Claude dashboards directly on existing Power BI semantic models, without opening Power BI Desktop or writing DAX. He covers step-by-step setup, how row-level security applies, sharing, and the current limitations.
  • Příprava dat pro AI: co nastavit v semantic modelu [Czech] by Jiří Neoral Jiří Neoral (in Czech) works through Power BI's Prep data for AI tools in order, from coarse to fine: trim the AI data schema to what Copilot should use, add verified answers for the most common questions, write focused AI instructions, and front-load descriptions, since Copilot reads only the first 200 characters. Only then does he mark the model Approved for Copilot. He flags the traps along the way: the schema isn't a security boundary and is ignored outside data Q&A, verified answers don't fully respect RLS/OLS or support CI/CD, and Czech questions proved unreliable in his tests.
  • Claude + Power BI: Build Dashboards on Your Semantic Model in Minutes (MCP Setup) by Atikant Jain Atikant Jain connects a Power BI semantic model in Fabric to Claude through the Power BI remote MCP server, then builds a Claude Dashboard on its governed measures. He covers the Entra app registration, redirect URI, API permissions and server registration, and closes with security, governance and how this compares with a Power BI report.
  • How to Use Direct Lake Across Public and Private Link Workspaces in Microsoft Fabric by Gilbert Quevauvilliers · 1 min read This article explains how to establish a Direct Lake connection between Public and Private Link workspaces in Microsoft Fabric, overcoming the previous limitation of not being able to link them directly to a Semantic Model.

Report Authoring and Interactivity

  • Project Controls Gantt 1.0: a free Gantt visual for Power BI by Juliana Smith · 2 min read Project Controls Gantt 1.0 enhances Power BI's Gantt visual by adding baselines, cut-off dates, dependencies, and critical task flags, supports up to 30,000 rows without DAX, is fully accessible, and is free under the MIT license.
  • Power BI Report CLI: What Changed for Accessibility? by Juliana Smith · 12 min read skills-for-fabric v0.3.17 merged Microsoft's four Power BI report skills into a single powerbi-report-cli skill. Juliana Smith finds the accessibility guidance is still there but only loads when a request routes through design mode, so quick edits and restyles that go straight to authoring skip it. She suggests reviewing for accessibility before fixing, spelling out requirements in prompts, and running a dedicated audit, and lists changes she'd like from Microsoft.
  • STOP Using SharePoint for Power BI Images - do this instead! by Patrick LeBlanc Patrick shows how to store Power BI images securely in a dynamic OneLake Lakehouse, removing dependence on SharePoint or anonymous blob storage using resilient GUID-based URLs and fine-grained security roles.
  • Choosing the Right Model In Fabric - Ep.575 - Power BI tips by Power BI Tips Mike and Tommy discuss how to match the right LLM to each Microsoft Fabric workload, asking whether bigger models are always better or just more expensive, and how teams can build a practical model-selection strategy across Power BI, notebooks, pipelines and more.
  • L’IA peut construire vos rapports Power BI… de A à Z (Partie 1) [French] by Audrée Pellerin · 1 min read Artificial intelligence can now build complete Power BI reports from scratch or modify existing ones by generating the necessary files, with the key enabling format being PBIP.
  • Diagrama de Gantt en Power BI: Nuevo visual nativo paso a paso [Spanish] by Guillermo Perdomo Sosa Learn how to create a native Gantt chart in Power BI using the new October 2026 preview visual, customizing task bars, dates, milestones, progress tracking, and formatting options step by step.
  • The Budget vs Actual Dashboard Every CFO Asks For by Chandeep Chhabra This Power BI video walks viewers through building Budget vs Actuals reports at three complexity levels, from simple monthly models to handling grain mismatches with bridge tables and creating rolling forecasts using DAX and conditional formatting.
  • Data Visualization in the Agentic Era by Alper Sarikaya · 8 min read Alper Sarikaya of the Power BI Core Visuals team introduces the Fabric visuals and Fabric data grid React libraries (preview), released alongside the Rayfin analytics template for Fabric Apps. Built on Vega-Lite so coding agents can generate charts from natural language, they borrow Power BI's visual style and add guardrails such as label truncation, minimum bar sizes, categorical scrolling and light/dark theming.
  • Power BI Portfolio Project: Build One With Claude (No Data) by Lukas Reese · 15 min read Lukas Reese builds a complete Power BI portfolio report from an empty PBIP file using Claude Code. Claude generates seasonal sample data, builds a star-schema model through the Power BI Modeling MCP server, and creates the report pages with the powerbi-report-cli skill after an HTML mockup is approved. He shares the exact prompts, the two Desktop settings to switch on, and seven lessons, including checking every measure Claude writes.

Deployment, Security and Operations

  • Copilot v Power BI: kde ho potkáte a z čeho se platí [Czech] by Jiří Neoral · 8 min read Jiří Neoral (in Czech) separates Copilot in Power BI into four features: the report Copilot pane, standalone Copilot, Copilot in apps, and Copilot for creators (DAX query view, measure descriptions, web modeling), plus Microsoft 365 Copilot. For each he covers where it runs, its limits and which capacity pays. He also explains F2/P1 requirements, Fabric Copilot capacity and token-based consumption, noting that a dozen questions used about a quarter of an F2's daily capacity in his test.
  • All you need is a Fabric app by Mimoune Djouallah · 7 min read Mimoune Djouallah introduces OneLake Studio, a free personal VS Code extension that browses OneLake, opens a Power BI semantic model (.bim) to explore it pivot-style, and compiles its DAX measures to SQL that DuckDB runs. It then publishes the model as a basic Fabric app that colleagues open in the browser. He is upfront about the limits: no RLS/CLS, small data only, possible egress costs and no guarantee of DAX fidelity.
  • Power BI Administration at Scale: PPCC26 Wrap-Up and Tenant Settings Demo Code by Heidi Hasting · 5 min read Following her PPCC26 session with Shabnam Watson, Heidi Hasting shares the demo code: a Fabric notebook that calls the List Tenant Settings admin API, keeps versioned snapshots and a change log in a lakehouse, and feeds a Direct Lake report showing whether each setting applies to the whole organisation or specific security groups. Deployment scripts build the whole kit in one command.
  • Distributed Wisdom for Fabric - Ep.576 - Power BI tips by Power BI Tips Mike and Tommy explore "distributed wisdom", connecting expertise spread across people and teams so it shapes decisions elsewhere, and whether semantic models, the OneLake catalog and AI agents can carry that judgement across an organisation or whether process and culture are the real blockers. They suggest practical first steps for BI teams, from governance guardrails to communities of practice.
  • The Govern Tab Shows Tenant Settings Now… Unless You’re Not an Admin by Heidi Hasting · 3 min read The OneLake catalog's Govern tab now shows tenant settings, embed codes, organisational visuals and more, but Heidi Hasting finds non-admins only see Capacities, Policies and Support (known issues and service health). She makes the case for a read-only view of tenant settings for everyone and, until then, suggests extracting settings via the admin APIs and sharing them in a report.
  • On-premises data gateway October 2026 release by Goutham Upadhyaya · 4 min read The October 2026 on-premises data gateway (3000.338, Power Query engine 2.159.229) aligns with the October Power BI Desktop release. It improves installer and configurator accessibility, keeps Power Query settings intact during connection tests, and gives better diagnostics for service-start failures. New Admin Connections REST APIs (preview) give Fabric admins tenant-wide visibility and control over every connection, and the post includes staged-upgrade best practices.
  • Validate every model change: deploy it right (part 2) by Ruben Van de Voorde · 1 min read A clean Git merge can still produce a broken semantic model. Part 2 of the series shows how to chain the Tabular Editor CLI's te validate (structural validity), Best Practice Analyzer (design rules) and pbir validate (downstream report fields) in a pipeline that uses exit codes to stop at the first failure, so a broken model is caught on every push.
  • Power BI Apps: Workspace Apps, Org Apps and Fabric Apps by Valerie Junk · 1 min read Workspace apps, org apps and Fabric Apps all appear on the Power BI Apps page. Valerie Junk explains what each one does and when to use which.
  • Unit Testing Semantic Models with DAX by Justin Martin · 1 min read A tool that runs YAML-defined DAX assertions against semantic models over the XMLA endpoint. For Direct Lake models, it deploys a temporary copy of the model for each pull request so the change can be tested before merging.
  • Power BI Git in VS Code: Track PBIP Changes by Ashish kumar Jha The first video in a Power BI with Git series explains why text-based PBIP suits source control better than binary PBIX. It then saves a report as a PBIP project, opens it in VS Code, and walks through git init, status, add, commit and diff to see exactly how a measure change shows up.

Get Power BI Weekly in your inbox every Friday

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.

Power BI Weekly Logo