--- title: FabConEU2026_PowerBI_Update_EN tags: author: [Eiki Sui](https://image.docswell.com/user/marshal_dabao) site: [Docswell](https://www.docswell.com/) thumbnail: https://bcdn.docswell.com/page/4EZL26KM73.jpg?width=480 description: FabConEU2026_PowerBI_Update_EN by Eiki Sui published: October 15, 26 canonical: https://image.docswell.com/s/marshal_dabao/5JWVQQ-2026-10-15-212833 --- # Page. 1 ![Page Image](https://bcdn.docswell.com/page/4EZL26KM73.jpg) Microsoft Fabric Power BI Microsoft Fabric | Power BI FabCon Europe 2026 Update Using data in the age of AI that “acts” For the CoE and business-unit Power BI users October 15, 2026 | Eiki Sui Focuses on Fabric / Power BI updates from Jan–Sep 2026 and FabCon Europe 2026 announcements (Sep 2026, Barcelona) | As of 2026-10-04 # Page. 2 ![Page Image](https://bcdn.docswell.com/page/Y76WV2Q57V.jpg) Introduction Today’s goals and how to read the tags Today’s goals 1 Get the big picture Where Microsoft invests (4 layers, Copilot) 2 See what changes in your work Changes for users, creators & the CoE 3 Decide your next actions What to try, configure, and migrate How to read the tags Availability GA Preview Available for production use For evaluation and testing; specs may change Coming soon Announced, not yet available Sneak peek Screens or concepts only (timing TBD) Roadmap Plans shared in sessions (timing TBD) Frontier Microsoft 365 Copilot early access Mixed Several statuses on one slide (shown per item) Action CoE decisions: settings, user comms, migration Audience CoE / Admins Creators Users V Video frame ★ Highlight 2 # Page. 3 ![Page Image](https://bcdn.docswell.com/page/G75MD2RG74.jpg) Introduction Agenda 01 The big picture of FabCon Europe 2026 02 Copilot and Fabric IQ 03 Semantic models in the AI era 04 Deep dive: storage modes Fabric overview, 4-layer stack, new Copilot Ask Power BI, delegate work, Fabric IQ Superpowers, Semantic views Evidence-based choice, true composite Everyone Users Creators / CoE Creators / CoE 05 Development in the agent era 06 Fabric Apps and Fabric Databases 07 Reports and distribution 08 For the CoE: admin and migration Developer Mode, spec-driven work with AI Creators Rayfin, Modern report builder, Pro/PPU Tables/matrices, web ribbon, sharing Deprecations, retirements, capacity, cost Everyone Creators / Users CoE The full deck also covers data preparation (Dataflow), Fabric IQ (Data Agent and Ontology), other updates (link list), and a glossary 3 # Page. 4 ![Page Image](https://bcdn.docswell.com/page/9J2984DDER.jpg) FabCon Europe 2026 Everyone FabCon Europe 2026: from the venue in Barcelona Event overview Dates September 28–October 1, 2026 Venue Barcelona, Spain Format FabCon (Fabric) + SQLCon (SQL), co-located The event in numbers ~5,000 130+ 10+ Data experts Sessions Tutorials + Community meetups and roundtables Atmosphere at the venue ■ A community openly sharing what it builds and learns ■ Theme: “The data foundation for Copilot and agents” Keynote venue (September 2026, Barcelona) FabCon and SQLCon 2026 in Barcelona (Azure Blog, 2026-09) FabCon Keynote @YouTube ■ The buzz mirrors Microsoft’s data product momentum 4 # Page. 5 ![Page Image](https://bcdn.docswell.com/page/DEY46M2MJM.jpg) Introduction Fabric and Power BI by the numbers 40,000+ 425,000+ Microsoft Fabric customers Power BI customers Worldwide In every country and industry 22M+ 15,000+ Leader Daily active Power BI semantic models Fabric App developers since the June 2026 launch Microsoft named a Leader in the Gartner Magic Quadrant The world’s most-used semantic layer Fabric Apps (Rayfin) Analytics and BI Platforms (2026) FabCon and SQLCon 2026 in Barcelona (Azure Blog, 2026-09) and other sources 5 # Page. 6 ![Page Image](https://bcdn.docswell.com/page/VJNYZW8378.jpg) 01 The big picture of FabCon Europe 2026 AI moves from “answering” to “acting.” First, let’s see where Microsoft is investing For Everyone Layer All 4 layers 6 # Page. 7 ![Page Image](https://bcdn.docswell.com/page/YE9PZXKPJ3.jpg) Microsoft Fabric overview Microsoft Fabric The unified data platform for AI transformation Fabric Items Data Factory Analytics Databases Real-Time Intelligence IQ Power BI Fabric Platform Copilot OneLake Governance 7 # Page. 8 ![Page Image](https://bcdn.docswell.com/page/GE8D12RXED.jpg) Microsoft Fabric | Data Factory (pipelines) Microsoft Fabric Data for Factory (Pipeline) The unified data platform AI transformation Connect to any data source with seamless, easy-to-use ETL • Extensive connectors for on-premises, Azure, and other cloud platforms • Intuitive, wizard-driven data copy experience • Support for custom code integration including SQL, Python, and .NET • Petabyte-scale data movement capabilities Data Factory Analytics Databases Real-Time Intelligence IQ Power BI Fabric Platform Copilot OneLake Governance 8 # Page. 9 ![Page Image](https://bcdn.docswell.com/page/LELMQ2KN7R.jpg) Microsoft Fabric | Data Factory (Dataflow Gen2) Microsoft Fabric The unified data platform for AI transformation Data Factory (Dataflow Gen2) Self-service Data Transformation • User-friendly experience built on Power Query experiences • 170+ supported data sources and 100+ built-in transformation functions • High-speed data movement for large-scale datasets with Fast Copy capability Data Factory Analytics Databases Real-Time Intelligence IQ Power BI Fabric Platform Copilot OneLake Governance 9 # Page. 10 ![Page Image](https://bcdn.docswell.com/page/4JMY182QJW.jpg) Microsoft Fabric | Analytics (notebooks) Data Engineering & Data Science (Notebook) Microsoft Fabric Multilingual Collaborative Notebooks Support for SQL, Python, Scala within a single notebook The unified• data platform for AI R, transformation Data Factory Analytics • Real-time collaboration with shared notebook editing capabilities • Low-code data transformation and profiling using Data Wrangler • AI-powered data transformations – summarize, extract, and translation • Automated AI model creation with AutoML • Custom function development using User Data Function (also accessible from Power BI) Databases IQ Power BI Real-Time Intelligence Fabric Platform Copilot OneLake Governance 10 # Page. 11 ![Page Image](https://bcdn.docswell.com/page/PJR9L5MK79.jpg) Microsoft Fabric | Databases Microsoft Fabric The unified data platform for AI transformation Databases Data Factory Analytics Databases SaaS-ified Databases • Operational data in Fabric • Native integration with OneLake • Reduced data duplication and movement • Built-in scalability, security, and governance IQ lower operational Poweroverhead BI • Real-Time Simplified architecture with Intelligence Fabric Platform Copilot OneLake Governance 11 # Page. 12 ![Page Image](https://bcdn.docswell.com/page/PEXQWKV5JX.jpg) Microsoft Fabric | Real-Time Intelligence Microsoft Fabric The unified data platform for AI transformation Real-Time Intelligence Analyze, visualize, and monitor data in real time • Seamlessly ingest and process streaming data such as IoT signals and application logs in real time • Monitor conditions continuously and trigger alerts via email or Microsoft Teams Databases Analytics Data Factory Real-Time Intelligence IQ Power BI Fabric Platform Copilot OneLake Governance 12 # Page. 13 ![Page Image](https://bcdn.docswell.com/page/3EK9V5QRED.jpg) Microsoft Fabric | Fabric IQ Microsoft Fabric The unified data platform for AI transformation Fabric IQ Unified Intelligence • AI-powered intelligence layer across the entire Fabric platform • Deep semantic understanding of your data estate Data Factory Analytics Databases Real-Time Intelligence IQ Power BI Fabric Platform Copilot OneLake Governance 13 # Page. 14 ![Page Image](https://bcdn.docswell.com/page/L73WMKPG75.jpg) Microsoft Fabric | Power BI Microsoft Fabric The unified data platform AI transformation Power BIfor (Report) No. 1 BI tool in the market • An intuitive, Excel-like user experience that anyone can use • Large-scale data analysis enabled by the new Direct Lake connectivity Data Factory Analytics Databases Real-Time Intelligence IQ Power BI Fabric Platform Copilot OneLake Governance 14 # Page. 15 ![Page Image](https://bcdn.docswell.com/page/87DK63LNJG.jpg) FabCon Europe 2026 highlights FabCon on one slide: 3 shifts Bringing governed data to where people and agents work 01 02 03 Agents move from “answering” to “acting” Compute (processing power) keeps up with agent demand Trust travels with the data ■ Fabric IQ is GA ■ GPU-accelerated queries ■ Sensitivity labels & DLP in Copilot in Copilot Chat and Cowork in Data Warehouse (preview) ■ Data engineering agent ■ On-demand billing: pay per use Related: p.22, p.23, p.42 Related: p.63 (preview) ■ Control via Restrict from Copilot ■ Note: Cowork doesn’t support DLP Related: p.30, p.31, full deck p.94 Power BI’s “next chapter”: Modern report builder, Semantic views, Apache Ossie, Developer Mode (→ p.19, full deck p.37, p.44) Bringing governed analytics into the flow of work (Fabric Blog, 2026-09) 15 # Page. 16 ![Page Image](https://bcdn.docswell.com/page/VJPKY42NE8.jpg) ★ Microsoft’s investment areas Top priority: 4-layer agentic analytics stack Conversational layer Semantic layer Serving layer Data foundation layer Copilot in Power BI / Data Agent Fabric IQ → Copilot Chat / Cowork Reaches people in the tools they use every day Semantic models / Semantic views (sneak peek) Ontology (preview) AI can reason over business logic Data Warehouse (GPU acceleration: preview) Cache control / Direct Lake Serves fast, to many users, with governance Data Engineering (Spark, Lakehouse) Data Science / Data Factory Keeps data fresh, scalable, and queryable Section 02, 08 Section 03, 08 Section 04 Section 07 Base: OneLake (open formats) | One security model | One governance | One “true definition” Building the agentic analytics stack (Fabric Blog, Build 2026) / Fabric Analytics at FabCon Europe 2026 “ AI does not merely assist, but executes across your data estate. AI doesn’t just assist; it takes action across your entire data estate 4 layers announced at Build 2026; FabCon Europe 2026 “completed the next step.” Key point: the 4 layers are designed as one 16 # Page. 17 ![Page Image](https://bcdn.docswell.com/page/2EVVGXDYEQ.jpg) ★ For every persona AI supports every persona’s work Built into the platform, AI supports everyone who uses, builds, and protects it Persona Key AI experiences (as of September 2026) Business users Copilot Chat and Cowork: ask Power BI, delegate work / Copilot in reports GA Analysts / report creators Create and summarize reports with Copilot, Fabric skill for report authoring, Modern visual defaults GA Modelers / BI developers Agentic development in Power BI (Authoring plugin), Copilot in web modeling, Remote Power BI Authoring MCP GA / Preview Data engineers Data engineering agent, Copilot in notebooks, Fabric Skills Preview / GA Admins / CoE Root cause analysis with Observability and Operations agent, Copilot in RealTime Intelligence App developers Fabric Apps (Rayfin) + GitHub Copilot, Fabric Core MCP FabCon and SQLCon 2026 in Barcelona (Azure Blog, 2026-09) Availability Preview Preview / GA 17 # Page. 18 ![Page Image](https://bcdn.docswell.com/page/57GL8VYWEL.jpg) ★ Microsoft-wide direction The new Copilot and Microsoft IQ: one front door The new Copilot One app to “ask, delegate, create, and automate” Home Office in Copilot Code Autopilot Chat and Cowork in one place Create and update Word, Excel, PowerPoint directly Anyone can create business apps & dashboards Always-on agents that work autonomously Frontier rollout Frontier rollout Frontier rollout Private preview Grounding (in your company’s context) Microsoft IQ The layer that connects company knowledge, data, and business context Work IQ How you work: email, meetings, chats, files Fabric IQ Business data and metrics (Power BI models, etc.) Foundry IQ Web IQ Policies, formal documents Information on the web Fabric IQ in Copilot Chat / Cowork: GA; integration with Code: Frontier (soon); Dynamics 365 / Power Platform data: preview Introducing the new Copilot (Official Microsoft Blog, 2026-09-25) / What is Fabric IQ? (Microsoft Learn) 18 # Page. 19 ![Page Image](https://bcdn.docswell.com/page/4EQY96MQJP.jpg) Power BI | Next chapter Everyone Power BI’s next chapter: from “view” to “act” Power BI evolves to power apps, agents, and Copilot on a trusted semantic layer 1 Coming soon 2 Coming soon 3 Sneak peek 4 Build with prompts Open to Pro / PPU Open up the semantic layer Copilot uses Power BI ■ Desktop’s ■ Build apps even ■ Semantic views ■ Ask Power BI from Modern report builder ■ Apps with input and write-back, too with Pro / PPU ■ Plus Fabric Database at no extra cost (sneak peek) ■ Driving Apache Ossie with Snowflake, etc. GA Chat and Cowork ■ Fabric IQ MCP is also GA For developers, “building with agents” becomes standard practice across Power BI and Fabric (→ Section 05) Power BI's next chapter (Power BI Blog, 2026-09) 19 # Page. 20 ![Page Image](https://bcdn.docswell.com/page/KJ4WX4VY71.jpg) 02 Copilot and Fabric IQ Use data in the flow of work Ask Copilot and get Power BI numbers back. Skip the report hunt and even hand off the work For Users Layer Conversational CoE / Admins 20 # Page. 21 ![Page Image](https://bcdn.docswell.com/page/LE1YN43N7G.jpg) Copilot and Fabric IQ Mixed Everyone Unifying Copilot: no more “Which one do I use?” “Chat with your data” moves to Microsoft 365 Copilot. Copilot in reports stays, answering the same way Before Now Power BI Power BI Standalone Copilot Copilot in reports Users Fabric Copilot per workload “Which one do I use?” Microsoft 365 Copilot Chat / Cowork Preview Power BI Standalone Copilot Gets the Microsoft 365 Copilot experience (differences → p.28) GA Copilot in Power BI reports Answers with the same engine as Copilot Chat (continues) Microsoft 365 Copilot Copilot Chat For creators: Copilot in Fabric (per workload) and agentic development Investment: Microsoft 365 Copilot & Copilot Cowork. Fewer doors, same governed answers anywhere Bringing Power BI Insights to Every Copilot User (Power BI Blog, 2026-07) 21 # Page. 22 ![Page Image](https://bcdn.docswell.com/page/GEWG4XXMJ2.jpg) Copilot and Fabric IQ GA Users Ask Copilot Chat about your Power BI data Overview If you can view a Power BI report, you can ask about it in Copilot Chat Copilot infers the report (or you specify it) GIF | Loops Demo videoautomatically | Autoplay, loop Key points ■ Answers via model measures, relationships, and business logic → the same numbers ■ No need to copy data or build AI-only models ■ RLS/OLS apply. So do Purview DLP and Information Protection ■ Put answers to work in email, meetings, docs Use cases “Last week’s sales?” “Where did KPIs drop?” Prerequisites Microsoft 365 Copilot license, etc. (→ full deck p.24) Microsoft 365 Copilot and Fabric IQ (Microsoft Learn, 2026-09) / FabCon Europe 2026 Power BI corenote (Microsoft, 2026-09) 22 # Page. 23 ![Page Image](https://bcdn.docswell.com/page/47ZL266MJ3.jpg) Copilot and Fabric IQ GA Users Copilot Cowork: delegate “from analysis to done” Demo video | Autoplay, loop ① What changes? Report insights become “finished work.” Summary emails, decks, invites in one flow ② Who benefits? Managers, sales planners, assistants. Routine report recaps, pre-meeting briefs ③ Considerations Usage-based billing (Copilot credits). Power BI reports and models only. No citations or DLP yet (labels are inherited). Check numbers in the report before sharing Example requests to Cowork “Email my boss last week’s shifts, next steps” “Add a report recap to Friday’s exec agenda” “If KPIs drop, draft a plan and book a review” Copilot Cowork and Fabric IQ (Microsoft Learn, 2026-09) 23 # Page. 24 ![Page Image](https://bcdn.docswell.com/page/YJ6WV225JV.jpg) Copilot and Fabric IQ Mixed Everyone The big picture of Copilot and agents: one meaning, wherever you ask Where you ask (entry) Note: Copilot in Fabric for creators (Data Factory, Data Engineering, Warehouse, Real-Time Intelligence, etc.) → full deck p.18 Copilot in Power BI Microsoft 365 Copilot Chat / Cowork Analysis in reports, apps & chat Insights in the flow of work Agents and apps Copilot Studio, Foundry, custom (via MCP) How Fabric grounds answers Note: availability details → p.26 Fabric IQ Ops intelligence GA Native to all layers of Fabric IQ Specialized agents spanning OneLake Ontology, entities, properties, actions Preview Business intelligence Unified data Fabric Data Agent The 3 layers → full deck p.70 Semantic models, reports, measures, KPIs Other sources SQL Database, Foundry Search, and more Governed structured and unstructured data in OneLake Governed foundation OneLake + Fabric governance Lakehouse Warehouse Eventhouse / KQL Semantic model Mirrored database Ontology Graph Shortcuts One governed foundation. Multiple conversational experiences. Consistent business meaning. FabCon Europe 2026 conversational analytics corenote (Microsoft, 2026-09) / What is Fabric IQ? (Microsoft Learn) 24 # Page. 25 ![Page Image](https://bcdn.docswell.com/page/GJ5MD22GJ4.jpg) Copilot and Fabric IQ | For newcomers Mixed Everyone Fabric IQ made simple: “tools” and “playbooks” for agents Fabric IQ = AI agents’ tools and playbooks (skills) for using company data correctly How Fabric IQ works From question to answer Copilot Chat Copilot Cowork GitHub Copilot Other agents Fabric IQ skills + tools 3 kinds of knowledge in Fabric IQ Ops intelligence Preview Ontology: how the business works Business intelligence Example: “Latest quarter’s sales by region?” 1 Find: reports/models with the answer 2 Read: structure, measures, descriptions 3 Check: find which column holds “Kanto” 4 Calculate: write DAX, run it as you (RLS and OLS still apply) 5 Explain: answer in words and tables GA ★ Power BI semantic models Unified data OneLake: Lakehouse, Warehouse, and more ★ = start here. Layers → full deck p.70 Skills Tools Step-by-step playbook for using tools Find, read, and run (read-only) What it means for Power BI users ① No need to rebuild Your existing semantic models become the AI’s “textbook” as is (→ p.27) ② The same wherever you ask Definitions, security, and governance reach every entry point together ③ 3 entry points, GA today (→ p.26) ■ Copilot Chat: instant answers ■ Cowork plugin: ask as you work ■ Fabric IQ MCP: build with agents Get started with the Fabric IQ MCP server (Microsoft Learn) / FabCon Europe 2026 conversational analytics corenote (Microsoft, 2026-09) 25 # Page. 26 ![Page Image](https://bcdn.docswell.com/page/9E29844D7R.jpg) Copilot and Fabric IQ Mixed CoE / Admins Creators Fabric IQ entry points: what’s here, what’s next Start with Power BI semantic models & reports (GA). Direct use of Data Agent/Ontology: preview coming soon Microsoft IQ: Work IQ (work) + Fabric IQ (business data) + Foundry IQ (policies, formal docs) + Web IQ (web) = company-wide context Semantic model Metrics, Power BI reports OneLake data Fabric IQ Ontology Structure and rules As context Preview soon Delivers meaning of business data Direct in Copilot Data Agent Q&A by business area Solid = GA Published as a Microsoft 365 agent Dashed = Preview, Frontier, upcoming Microsoft 365 Copilot (Chat and Cowork) GA Copilot in Power BI (in reports) GA GitHub Copilot and other MCP clients GA Copilot Code tie-in (soon) Frontier Copilot Studio/Copilot Chat, etc. GA Note: Cowork currently supports only Power BI reports and models (not Data Agent or Ontology) What is Fabric IQ? (Microsoft Learn) / FabCon Europe 2026 session W23 (Microsoft, 2026-09) 26 # Page. 27 ![Page Image](https://bcdn.docswell.com/page/D7Y46MMMEM.jpg) Copilot and Fabric IQ GA Everyone Conversational analytics with Fabric IQ: the work you’ve already done pays off What agents need is already in semantic models. Same numbers, same governance, now conversational What your semantic model already has What agents need DAX measures and business logic Trusted KPI definitions (AI doesn’t rebuild) Relationships, star schema, hierarchies Correct joins (AI doesn’t guess) Clear names, formatting, synonyms Vocabulary that matches how users ask RLS, OLS, sensitivity labels Per-user security applied automatically Promoted / certified (content endorsement) Clear signal of which models to trust No need to rebuild: your investment in existing models directly drives answer quality (→ p.35) FabCon Europe 2026 session W23 (Microsoft, 2026-09) 27 # Page. 28 ![Page Image](https://bcdn.docswell.com/page/VENYZWW3J8.jpg) Copilot and Fabric IQ Mixed Users CoE / Admins Two Copilot × Power BI experiences compared Use Power BI in Copilot GA Use Copilot in Power BI Preview Power BI data answers in Microsoft 365 Copilot Chat Microsoft 365 Copilot in Power BI (Copilot Premium license) What it uses Power BI reports and models (artifacts) Fabric IQ (data) + Work IQ (email, meetings, files) What you can do Ask, spot trends (Chat); create documents (Cowork) Share chat history across apps, start Cowork tasks License / setup Microsoft 365 Copilot (Premium) license Admin settings, cross-geo processing → full deck p.24 Microsoft 365 Copilot license (early preview) Enable via new Fabric admin setting (off by default) Without a Microsoft 365 Copilot (Premium) license ■ Copilot in Power BI (standalone, in reports) works as before: on Fabric capacity (F2 or higher) or Fabric Copilot Capacity ■ Report Copilot with Fabric IQ: transparent multi-step reasoning, visual generation, DAX/data table review (available now) ■ A unified Copilot for Fabric users without a license is also planned (capabilities may be limited) Power BI September 2026 Feature Summary (2026-09) / FabCon Europe 2026 Power BI corenote (Microsoft, 2026-09) 28 # Page. 29 ![Page Image](https://bcdn.docswell.com/page/Y79PZXXPE3.jpg) Copilot and Fabric IQ | Sneak peek Sneak peek Users Creators Fabric IQ from Copilot in Excel: PivotTables through chat Overview Demo video | Autoplay, loop Just ask Copilot in Excel to analyze Power BI semantic models Key points ■ Create dynamic PivotTables on semantic models through chat ■ Copilot finds the right model and the best fields and measures ■ Follow-up questions, new charts, and analysis updates, all in chat Note Presented in a FabCon session (planned; subject to change) FabCon Europe 2026 conversational analytics corenote (Microsoft, 2026-09) / FabCon Europe 2026 session W23 (Microsoft, 2026-09) 29 # Page. 30 ![Page Image](https://bcdn.docswell.com/page/G78D122X7D.jpg) Copilot and Fabric IQ Mixed CoE / Admins CoE governance: decide before opening up to AI Access control: which models to open to AI Data protection: how to secure AI answers ■ Restrict from Copilot: AI control per model → p.31 ■ Sensitivity labels: chats and outputs inherit ■ Approved for Copilot: flag recommended models to surface them in search (formerly Prepped for AI) the most restrictive label (Chat and Cowork) ■ DLP: applies in Chat; Cowork not supported yet ■ RLS/OLS: enforced for each user, even via AI ■ Information Protection: chat export on/off (Chat) ■ Before publishing, complete AI prep such as ■ Purview DSPM for AI: monitor and protect descriptions and synonyms (→ full deck p.38) Copilot and Data Agent usage (preview) Note: prerequisites (licensing, tenant settings, cross-geo processing) and billing → full deck p.24 Action Before company-wide rollout, decide: AI model inventory, label & DLP operations, Cowork scope Fabric IQ in Microsoft 365 Copilot Chat (Microsoft Learn) / FabCon Europe 2026 session W23 (Microsoft, 2026-09) 30 # Page. 31 ![Page Image](https://bcdn.docswell.com/page/L7LMQ22NJR.jpg) Copilot and Fabric IQ GA Creators CoE / Admins Restrict from Copilot: control AI use per model Overview Model editors control read-only users’ AI use (Copilot, MCP) per model → You can open only approved models to AI GIF | Loops Demo videoautomatically | Autoplay, loop Where to set it Model settings > Copilot; on by default Experiences covered ■ Copilot in Power BI ■ Microsoft 365 Copilot (Chat and Cowork) ■ Data Agent, Microsoft’s remote MCP When turned off ■ Read-only users can’t reach the model ■ Hidden from all searches (editors: direct use) ■ Setting is not inherited (set per model) Use Copilot with semantic models in Power BI (Microsoft Learn) 31 # Page. 32 ![Page Image](https://bcdn.docswell.com/page/4EMY188QEW.jpg) 03 Semantic models in the AI era For Copilot, Data Agent, and apps alike, the “semantic layer” determines answer quality For Creators Layer Semantic CoE / Admins 32 # Page. 33 ![Page Image](https://bcdn.docswell.com/page/PER9L55KJ9.jpg) ★ Semantic models in the AI era Everyone The first AI asset to build is a “trusted semantic model” In an AI world, semantic models are the top investment The reasons: 5 strengths Microsoft highlights (why they help agents → p.27) 1 Company-wide metrics & KPIs defined in one place “Sales” and “gross profit” aligned company-wide. No gaps across reports 4 Business-friendly names and inferred relationships Work with data in business terms, not technical table and column names 2 Foundation for consistent, trusted AI answers Copilot and Data Agent alike query the model with DAX to answer 5 Industry leader & outperformer 22M models, 38M+ monthly users Used by 94% of the Fortune 500 3 Mature APIs and open-standard file formats XMLA, REST API, TMDL, and more, usable from tools and agents “It’s tempting to assume AI makes the data model less important. The opposite is true.” — Power BI Blog (2026-07) FabCon Europe 2026 Power BI corenote (Microsoft, 2026-09) / Bringing Power BI Insights to Every Copilot User (Power BI Blog, 2026-07) 33 # Page. 34 ![Page Image](https://bcdn.docswell.com/page/P7XQWKK5EX.jpg) ★ Semantic models in the AI era Everyone One model, many uses Power BI reports Fabric IQ Ontology Visualize and share Imports DAX measures as metrics Analyze in Excel Fabric Apps Analyze with PivotTables Connector DAX queries (read-only) Microsoft 365 Copilot Approved Semantic model Answers in Chat and Cowork Single source of truth for metrics Planning Plan vs. actuals Fabric Data Agent GitHub Copilot, MCP, LLMs When the semantic model is the source Development and automation Note: many of these also use other sources (SQL, KQL, etc.). Shown here: the “semantic model as single source of truth for metrics” setup Model investment pays off with every use. Definitions set once yield the same numbers everywhere Compiled for this deck (based on each feature’s official documentation) 34 # Page. 35 ![Page Image](https://bcdn.docswell.com/page/37K9V55R7D.jpg) Semantic models in the AI era Everyone What is a semantic model? How it differs from a metrics layer Not just a metrics store: governed security, time intelligence, and a full calculation engine built in Typical metrics layer Metrics and dimensions Basic joins Power BI semantic models RLS/OLS Time intelligence Custom/fiscal calendars Row/column-level security YTD, YoY, moving averages Fiscal periods like 4-4-5 Ragged & parent-child Calculation groups/UDFs Full DAX calc engine Org & account hierarchies Reusable logic Complex calcs in the model Many-to-many & bi-di Parameters Synonyms Expressive relationships Field / What-if AI understands user terms Simple row filters SQL aggregations only What it lacks Governed security Time intelligence Full calculation engine FabCon Europe 2026 Power BI corenote (Microsoft, 2026-09) / Semantic models in Power BI (Microsoft Learn) 35 # Page. 36 ![Page Image](https://bcdn.docswell.com/page/LJ3WMKKGJ5.jpg) Semantic models in the AI era Sneak peek Creators CoE / Admins Semantic views: define once, reuse everywhere Overview Demo video | Autoplay, Click to play loop (with audio) A way to define metrics and meaning where the data lives (OneLake) Key points ■ Manage AI context with the data, too ■ Share governed metrics across teams ■ Open standard: vendor-neutral, usable across platforms (→ full deck p.37) Demo ① Lakehouse: define metrics in YAML ② Shortcut other teams’ definitions ③ Same numbers in model, report, Copilot FabCon Europe 2026 Power BI corenote (Microsoft, 2026-09) / Power BI's next chapter (Power BI Blog, 2026-09) 36 # Page. 37 ![Page Image](https://bcdn.docswell.com/page/8JDK633NEG.jpg) 04 Deep dive Choosing a storage mode Report speed directly drives user satisfaction. Choose based on “evidence” For Creators CoE / Admins Layer Serving Semantic Note Deep dive (as needed) 37 # Page. 38 ![Page Image](https://bcdn.docswell.com/page/VEPKY44N78.jpg) Deep dive: storage modes Creators CoE / Admins Storage mode options and characteristics Mode Data location Strengths Considerations Best for Import Model memory Fastest, rich modeling features Requires refresh, has size limits Up to mid-size, complex calculations DirectQuery Source (e.g., Databricks SQL warehouse) Latest data, no data movement Depends on source performance, concurrency Real-time required, data can’t be moved Direct Lake on OneLake Delta in OneLake. Needed parts into memory No refresh, near-Import speed. Multiple sources Row limits per capacity. Delta layout matters too Data can sit in Fabric, mid-to-large scale Direct Lake on SQL Via the SQL endpoint DirectQuery Fallback available Performance on fallback SQL endpoint-centric existing setups Older composite + aggregations Dual/DirectQuery + Import aggregations Aggregation hits: fastest and stable Aggregation misses: source performance Very large scale with predictable queries Direct Lake composite+Import Delta from multiple items + Import tables Regular relationships, full Direct Lake speed Preview. Authoring mainly in web modeling Fabric data plus small master & external data No one mode fits all. Choose by data location, scale, freshness, query patterns (→ p.39, p.40) Direct Lake overview (Microsoft Learn) 38 # Page. 39 ![Page Image](https://bcdn.docswell.com/page/27VVGXXY7Q.jpg) Deep dive: storage modes Mixed Creators CoE / Admins Direct Lake on OneLake’s “true composite model” ① Direct Lake on SQL Lakehouse SQL analytics endpoint Semantic model GA Create in SQL analytics endpoint ② Direct Lake on OneLake Lakehouse Direct read: OneLake Delta Create in Lakehouse Direct Lake on SQL Single data source Lakehouse tables Preview Sales Inventory Inventory dim_Calendar dim_Product Warehouse tables Aspect Older composite DirectQuery+Import Direct Lake composite Direct Lake+Import Joins Limited Cross-source joins Regular Same-engine joins Speed Extra work per join often slows queries Same as Direct Lake (runs in memory) Model DAX: no blank row, no table expansion Hierarchies, calc columns on Import dim_Customer Tables switched to Import dim_Calendar Multiple sources No fallback Report GA Multiple sources + Import: true composite model Sales dim_Product Semantic model or in Power BI Desktop (OneLake catalog) Direct Lake on OneLake Lakehouse tables Single source Fallback available Report Multi-source + Import in one model, regular relationships = far more modeling flexibility. A Microsoft edge today Deep dive into composite semantic models with Direct Lake and import tables (Power BI Blog, 2026-02) / Strong and weak relationships in Power BI (SQLBI) 39 # Page. 40 ![Page Image](https://bcdn.docswell.com/page/5JGL8VVW7L.jpg) Deep dive: storage modes Creators CoE / Admins Decision flow and practical tips When data freshness & query speed matter (guide) Q1 Current Import or similar: fast and fresh enough? Yes Per product team Keep it as is Switching isn’t required No: improve freshness & speed Q2 Can your data live in OneLake? Lakehouse Yes First choice: Direct Lake on OneLake Yes Composite + aggregations Fit aggregations to real usage Warehouse No Q3 Billions of rows, with predictable top queries? Real-time required, or data can’t be moved? Yes DirectQuery Size and cache the source, too No Improve within Import Incremental refresh, model tuning Modern Power BI Architecture Choices (white paper, 2026-09) ■ Always test with your data & queries. Measure Cold/Warm/Hot separately ■ Delta layout affects performance. Design V-Order, Liquid Clustering, and row group size ■ Capacity and SQL warehouse sizes change the results ■ DirectQuery / Direct Lake: No Q4 Practical tips expertise in both Power BI and Databricks / Delta is required ■ Results as of Jun–Jul 2026. Both products are improving fast 40 # Page. 41 ![Page Image](https://bcdn.docswell.com/page/47QY966QEP.jpg) 05 Power BI development in the agent era Agents build; you focus on specs and validation. What’s “normal” in development is changing For Creators CoE / Admins 41 # Page. 42 ![Page Image](https://bcdn.docswell.com/page/KE4WX44YJ1.jpg) Agentic development | The new standard Mixed Creators Agentic Development becomes the new “normal” For developers, building with agents becomes the standard across Power BI & Fabric Data ingestion Greenfield Building new Brownfield Improving existing Transform New requirements Model design Identify gaps Reports & apps Improve Models, reports, apps GA Power BI Agentic Experiences Either way: delegate to agents → review results → improve (don’t build from scratch) Expanding across Fabric Data engineering agent Preview Plans, runs, validates data engineer tasks (hours–days) Fabric Skills Public on GitHub Skills for GitHub Copilot, Claude, and CLI Power BI September 2026 Feature Summary (Power BI Blog, 2026-09) Fabric Core MCP GA AI agents operate Fabric safely (cloud-hosted) Fabric IQ MCP GA Business context to external AI clients 42 # Page. 43 ![Page Image](https://bcdn.docswell.com/page/L71YN44NJG.jpg) Agentic development | Toolkit Mixed Creators Toolkit overview: agent + plugins + MCP Input Agent Plugins Goal / ask powerbi-authoring What to build / fix + Context Requirements, refs, standards Validate GitHub Copilot App Planning, tool calls, building Power BI Authoring MCP servers End-to-end model creation handled by AI agents Semantic modeling and report authoring skills and tools (CLI, MCP), all in one Rayfin Work IQ Preview Build a Fabric App (→ full deck p.56) Ships with the plugin Edits Desktop / PBIP Desktop Bridge Reloads code changes, then previews results GA Email, meeting, and file context Local MCP Agentic development in Power BI (Microsoft Learn) GA Model / report / app PBIP: GA Check in browser Confirms remote report edits Remote Power BI Authoring MCP or Output Developer Mode → p.44 Preview Edits models in Fabric, no install needed 43 # Page. 44 ![Page Image](https://bcdn.docswell.com/page/G7WG4XZME2.jpg) Agentic development | Developer Mode Mixed Creators Power BI Developer Mode: reports and models as code Overview Demo video | Autoplay, Click to play loop (with audio) PBIP and PBIR are GA (September 2026) Treat reports and models as “code” Key points ■ Ready for Git, CI/CD, and AI editing ■ PBIR: the default report format ■ New: Desktop detects external edits → Just apply to update instantly (preview) ■ “Open in VS Code” (preview) How to start Under Preview features, turn on: Detect and reload external PBIP changes (Desktop August 2026 or later) External editing of Power BI projects (Microsoft Learn) / FabCon Europe 2026 Power BI Corenote (Microsoft, 2026-09) 44 # Page. 45 ![Page Image](https://bcdn.docswell.com/page/4JZL261ME3.jpg) Agentic development | Workflow Creators Vibe or Spec: the spec-driven approach Vibe (ad hoc prompts) Spec (spec-driven) Loose prompt → eyeball it → keep tweaking Fast Exploratory → Hard to repeat Spec → plan/split → build → verify vs. spec Predictable Reviewable Repeatable Spec-driven Agentic Loop Colored = human-led steps 1 Plan 2 Review 3 Implement 4 Validate 5 Refine Plan document (.md) Vet assumptions, gaps Build with tools Run, test, diff Update plan with gaps Missing assumptions accumulate in the plan → next requests get shorter GitHub Copilot App agent modes: Interactive / Plan / Autopilot mode Typical flow = agree in Plan → build in Autopilot mode. Distinct from Microsoft Copilot’s “Autopilot” (p.18) FabCon Europe 2026 session W34 (Microsoft, 2026-09) 45 # Page. 46 ![Page Image](https://bcdn.docswell.com/page/YE6WV2L5EV.jpg) 06 Fabric Apps and Fabric Databases Beyond reports: the last mile linking insights to “decide, act, record.” Open to Pro / PPU too For Everyone Layer Semantic → apps 46 # Page. 47 ![Page Image](https://bcdn.docswell.com/page/GE5MD21GE4.jpg) Fabric Apps | Overview Preview Everyone What are Fabric Apps? From “insight” to “action” An app layer linking report insights directly to business decisions and actions Understand (reports) Decide Act Record (apps) 15,000+ Fabric App developers since June 2026 Makers now include designers and finance Fabric App Rayfin Fabric-hosted web apps Built on Fabric data Runs on Rayfin a code-first framework (how it works: p.48) Semantic model SQL database Data it uses V16 | Video in preparation Fabric App demo (input → save → update) 20 sec, autoplay/loop (delete after adding video) Warehouse From prompt to production: What’s new in Fabric Apps (Fabric Blog, 2026-09) Lakehouse 47 # Page. 48 ![Page Image](https://bcdn.docswell.com/page/9729841DJR.jpg) Fabric Apps | Rayfin Preview Creators CoE / Admins What is Rayfin? A framework for building Fabric Apps in code 1 Write 2 You or a coding agent writes the app in TypeScript Generates typed APIs and clients Runs the whole app locally GitHub Copilot + Rayfin plugin Building from chat → full deck p.56 Generate & run locally 3 Deploys as a Fabric App No infra tickets/secret sprawl Rayfin SDK / CLI Fabric runs/governs Open source (microsoft/rayfin) DB, storage, connector, function, auth One project, everything in code UI Data schema Connections Functions Auth config TypeScript rayfin/data rayfin.yml rayfin/functions Defined in code Everything is typed code → field names and calls are explicit Agents work from the app’s actual model, not guesswork From prompt to production: What’s new in Fabric Apps (Fabric Blog, 2026-09) Deploy rayfin up What deployment does ■ Provisions services ■ Applies schema ■ Deploys functions Data stays where it is 48 # Page. 49 ![Page Image](https://bcdn.docswell.com/page/DJY46MZM7M.jpg) Fabric Apps | Modern report builder Coming soon Creators Users Modern report builder in Power BI Desktop Overview New authoring experience in Power BI Desktop Instead of a blank canvas, AI builds a data app in seconds Demo video | Autoplay, Click to play loop (with audio) Key points ■ BI with prompts Describe the dashboard you need and iterate ■ Build in your usual tools Publish straight to Fabric ■ Go further with Fabric Apps Input, data write-back, workflows Use cases Sales deal tracker (log comments and next actions) Budget request/approval, inventory allocation FabCon Europe 2026 Power BI Corenote (Microsoft, 2026-09) / Power BI’s next chapter (Power BI Blog, 2026-09) 49 # Page. 50 ![Page Image](https://bcdn.docswell.com/page/V7NYZW33E8.jpg) Fabric Apps | Open to Pro / PPU Coming soon Everyone Open to Pro / PPU: Fabric Apps & Fabric Database At no extra cost, Power BI Pro / PPU customers also get Fabric Apps and Fabric Database capabilities (up to 1 GB per app) Power BI Pro / PPU No extra cost Fabric Apps + Fabric Database Up to 1 GB per app For business units For the CoE ■ Build apps beyond reports, no capacity needed ■ Governance for app growth (full deck p.61) ■ Analytics and operational data (Fabric Database) in one ■ Confirm details (limits, regions) at launch app ■ Build with AI (natural language) FabCon and SQLCon 2026 in Barcelona (Azure Blog, 2026-09) ■ With Fabric capacity, extend by combining with other workloads 50 # Page. 51 ![Page Image](https://bcdn.docswell.com/page/YJ9PZX9P73.jpg) 07 Reports and distribution Build better reports faster; view where you work For Creators Users 51 # Page. 52 ![Page Image](https://bcdn.docswell.com/page/GJ8D129XJD.jpg) Reports and distribution | Distribution channels Mixed Users Distribution channels move into the flow of work Teams GA PowerPoint Coming soon Tabs in private and shared channels, too Item info header at the top of tabs (GA) Link cards: updated date, sensitivity label (coming soon) Embed live pages and visuals Tell the story right in your slides Also supports reports in Org Apps Outlook Excel Preview Share live visuals in emails in Outlook on the web How to use it: next slide (→ p.53) SharePoint Power BI Latest numbers where you are Preview Enhanced embedding in SharePoint Online Web part: no need to paste a URL, just select a workspace and report Power BI September 2026 Feature Summary (Power BI Blog, 2026-09) Not stated GA Higher limit for export to Excel 150,000 → 500,000 rows (GA) Export from mobile apps too OneDrive / SharePoint Not stated Save and share from Power BI Desktop Version history in Desktop Note: legacy file picker retired (→ p.61) 52 # Page. 53 ![Page Image](https://bcdn.docswell.com/page/LJLMQ2WNER.jpg) Reports and distribution | Outlook Preview Users Share “live” visuals in Outlook Overview Demo video | Autoplay, loop Share interactive Power BI visuals in emails in Outlook on the web Key points ■ Just pick a visual in Power BI and share ■ Unlike screenshots, never goes stale ■ Loads in seconds; explore in the email ■ Also available in Loop Use cases Share weekly numbers with your team. Email your manager the latest charts Power BI September 2026 Feature Summary (Power BI Blog, 2026-09) 53 # Page. 54 ![Page Image](https://bcdn.docswell.com/page/47MY189Q7W.jpg) Reports and distribution | Org Apps GA Creators CoE / Admins Org Apps: distribution tailored to each audience Overview Org Apps Demo video | Autoplay, loop Build multiple apps per workspace, and tailor content for each audience Key points ■ Audiences (GA) also work on mobile ■ Bookmarks, comments, use in PowerPoint ■ Distribute Fabric items like notebooks ■ Supports CRUD APIs, pipelines, and Git ■ Fabric Planning can be added too ■ Fabric Apps can be added (coming soon) Use cases One org-wide app, tailored by department Org apps with audiences for Power BI and Fabric (Power BI Blog, 2026-06) V17 | Video in preparation Switching audiences changes what you see (optional) 10–15 sec (GIF), autoplay/loop (delete after adding video) 54 # Page. 55 ![Page Image](https://bcdn.docswell.com/page/P7R9L5GKE9.jpg) Reports and distribution | Core visuals Mixed Creators Users Core visuals: key updates in 2026 Power BI's built-in visuals evolve across look, interaction, and analytics 1 Style it all at once → p.56 2 Slicers ★ Modern visual defaults GA · Aug ★ Date picker ★ Design ribbon GA · Sep ■ Input slicer: numbers ■ “Versa” palette, Dark theme ■ Sample reports redesigned too 4 Calculations & charts → p.57 GA · Aug GA · Feb, May ■ Paste values into any slicer Sep → full deck p.85 ■ Slider colors 5 Cards & tooltips Feb GA · Sep → full deck p.85, p.86 3 Table and matrix → p.58 ★ +/- on column headers GA · Aug ■ Auto-expand row/col headers GA · Jun ■ Fix/reset column widths GA · May, Sep ■ Row headers frozen by default Aug 6 Maps & narratives → full deck p.85, p.86 ★ Visual calcs, custom totals GA · May ■ Card: highlight selected category Apr ■ Azure Maps markers ■ Calcs in small multiples too GA · Sep ■ Card: fixed-size layout Apr ■ Azure Maps auto-zoom GA · Aug ■ Line/legend cond. formatting GA · Jul ■ Card: image hover GA · Jun ■ Azure Maps geocoding GA · Sep ■ Initial axis scroll position GA · Sep ■ Modern tooltips GA · Jan ■ Shape map GA · Jun ■ Donut center value GA · Aug ■ Tooltip options GA · Jun ■ Narrative: select dynamic values Jan Sep ★ = biggest impact on business users (this deck's view). Right: status and Feature Summary month (month only = no status in its heading) Power BI Feature Summary (Jan–Sep 2026) 55 # Page. 56 ![Page Image](https://bcdn.docswell.com/page/PJXQWKX57X.jpg) Reports and distribution | Modern visual defaults and Design ribbon GA Creators Style the entire report at once Before Now → Still image | Video in preparation Previous format pane (per-visual settings) V12 | Video in preparation Change theme in Design ribbon; everything updates Still image, autoplay/loop (delete after adding video) 15–20 sec, autoplay/loop (delete after adding video) Format visual by visual; consistency takes effort Power BI September 2026 Feature Summary (Power BI Blog, 2026-09) Polished by default; Design ribbon styles it all 56 # Page. 57 ![Page Image](https://bcdn.docswell.com/page/3JK9V5WRJD.jpg) Reports and distribution | Slicers GA Creators Users Slicers and date selection Overview Demo video | Autoplay, loop Date picker: choose periods on a calendar Input slicer: filter by typing or pasting Key points ■ Date picker: relative & single date ■ Input slicer: cond. formatting, numbers ■ Paste selected values into any slicer ■ Set slider color and shape Use cases Select “last 3 months” right away. Paste an Excel list to filter in one go Power BI August 2026 Feature Summary (Power BI Blog, 2026-08) V13 | Video in preparation Select “last 3 months” in the date picker 10 sec (GIF), autoplay/loop (delete after adding video) 57 # Page. 58 ![Page Image](https://bcdn.docswell.com/page/LE3WMK1GE5.jpg) Reports and distribution | Table and matrix Mixed Creators Users Table and matrix reach the next level Overview Demo video | Autoplay, loop Table and matrix upgrades now GA. Even big tables are easy to read and use Key points ■ Auto-expanding row/column headers let you pivot the way you want ■ Fixed widths keep the layout intact even when the data changes ■ Filter and sort right on the visual (coming soon) Long-awaited Column header +/- was a top-3 request on Power BI Ideas. Finally GA Power BI August 2026 Feature Summary (Power BI Blog, 2026-08) 58 # Page. 59 ![Page Image](https://bcdn.docswell.com/page/8EDK63XN7G.jpg) Reports and distribution | Web authoring Preview Creators CoE / Admins Report ribbon on the web (preview) Overview Demo video | Autoplay, loop The full report ribbon on the web too, in both view and edit modes Key points ■ Organized in tabs, just like Desktop ■ Format painter, Optimize ribbon, and Design ribbon (→ p.56) come to the web ■ Switch with “Try the new ribbon” (turn it off to go back) For the CoE With web modeling (→ full deck p.39), the “browser-only authoring” trend grows. Review how widely to allow web editing Power BI September 2026 Feature Summary (Power BI Blog, 2026-09) 59 # Page. 60 ![Page Image](https://bcdn.docswell.com/page/V7PKY4PNJ8.jpg) 08 CoE: administration, governance, and migration First pin down what stops and what changes, then decide the rollout order For CoE / Admins 60 # Page. 61 ![Page Image](https://bcdn.docswell.com/page/2JVVGX2YJQ.jpg) For the CoE | Deprecations and retirements Action Everyone Deprecations and retirements (Power BI) 1/2 Item (link) Timing Impact Action Related links Power BI Q&A retirement (incl. Q&A setup tool) Feb 2027 (delayed from Dec 2026) Q&A visuals are removed and show errors. Embedded Q&A has no direct replacement Move to Copilot (F2 and up). Inventory where it's used Delay notice Learn: Copilot Power BI Desktop legacy file picker retirement From Oct 2026 (old UI retired in Apr 2026) Desktop Mar 2026 or older can't save or share to OneDrive or SharePoint Update Desktop to the latest (deployment packages too) New file picker Legacy “Excel/CSV import” deprecation in the service Announced Feb 2026 Only the old import UI retires. Still fine as a data source Use the Create page in the service, or Desktop Detailed article Learn Scorecard “hierarchy” removal Apr 15, 2026 Scorecard views auto-built from hierarchies stop working Check the impact and consider alternative views Learn: Scorecards End of support for SSRS Report Viewer web part for SharePoint Apr 13, 2026 Download still available; no more updates or support Find where it's used and consider alternative displays Related links Item names link to the relevant Feature Summary section. Detailed impact, actions, and all links are in the speaker notes Power BI Feature Summary (Jan–Sep 2026) 61 # Page. 62 ![Page Image](https://bcdn.docswell.com/page/5EGL8VRWJL.jpg) For the CoE | Deprecations, retirements, and related changes Action Everyone Deprecations & retirements 2/2 + related changes Item (link) Timing Impact Action Related links SCOM mgmt pack deprecation Announced Feb 2026 SSRS, PBIRS & SSAS monitoring Review how you monitor SQL Server Blog Simba Vertica ODBC deprecation From Feb 2026 Driver built into the Vertica connector Confirm switch to new driver Learn: Vertica QuickBooks Online retirement Mar 2026 Connector no longer supported Check usage & alternatives — Netezza in-box ODBC deprecation From Apr 2026 To new IBM Netezza ODBC driver Validate with the new driver Learn: Netezza Related changes (link) Timing Details and action Dataflow Gen1 moves to Legacy Apr 2026 No new features; 12+ mo. retirement notice. Plan Gen2 move (full deck p.63, p.64) ODBC to ADBC migration From Oct 2026 Planned (through spring 2027). Hive deprecated with no replacement. Inventory and validate with pq-adbc-advisor (full deck p.93) Data Agent in Copilot retired Aug 26, 2026 Use Data Agent via M365 Copilot, Copilot Studio, Foundry, or MCP Data Agent API integration Aug 26, 2026 With the Assistants API ending, move custom calls to the new API Narrative visual dynamic values Sep 2026 Moves from Q&A to a new selection UI. Update authoring steps (full deck p.86) Subscription email delivery From Sep 2026 Phased delivery platform change. New sender/format: check allowlists, inbox rules PBIR: the default report format Jan & Jun 2026 Default in service, then Desktop. Check internal tools and procedures for PBIR Power BI / Fabric Feature Summary (Jan–Sep 2026) 62 # Page. 63 ![Page Image](https://bcdn.docswell.com/page/4JQY96VQ7P.jpg) For the CoE | Capacity and cost Mixed CoE / Admins New options for capacity and cost Item (link) Details Status Main audience Capacity for Copilot Copilot runs on Fabric capacity F2 and up. Fabric Copilot Capacity can be created from F2 too GA Pro / PPU-centric organizations On-demand billing / Zero-provisioned (F0) Pay only for what you use, no reserved capacity (to be offered as a preview) Capacity overage Flexibility to exceed capacity at peak times Preview Busy-season teams Surge protection (per workspace) Stop a few jobs from overloading capacity Preview CoE Fabric Apps & Fabric Database on Pro / PPU Up to 1 GB per app, no extra cost Copilot Cowork Usage-based billing (Copilot credits). Manage budgets with Agent 365 Coming soon Starting small Coming soon Business units GA Users / CoE Pricing and availability may vary by contract and region. For the latest, see links in the speaker notes Build, deploy, and govern Microsoft Fabric at scale (Fabric Blog, 2026-09) 63 # Page. 64 ![Page Image](https://bcdn.docswell.com/page/K74WX4MYE1.jpg) Summary Summary: 3 key takeaways from today 1 AI moves from “answering” to “acting” 2 Answer quality depends on the semantic model 3 Beyond reports: the way we build changes too Copilot becomes one front door. Ask about your Power BI numbers and even hand off the work (Fabric IQ in Copilot Chat / Cowork: GA) Copilot answers using the Power BI semantic model, and Data Agent, apps, and Ontology can reuse it as the single source of truth for metrics. Models pay off most Build apps even on Pro / PPU, and develop together with agents. Meanwhile, plan ahead for what “stops”: Gen1, Q&A, ODBC, and more Next step: Pick your team's “top 10 questions” and “the model that answers them” 64 # Page. 65 ![Page Image](https://bcdn.docswell.com/page/LJ1YN48NEG.jpg) Summary | Reference links Everyone Reference links FabCon & SQLCon 2026 announcements The next chapter for Power BI Azure Blog, 2026-09 Power BI Blog, 2026-09 Fabric Analytics at FabCon Europe Bringing Power BI insights to Copilot users Fabric Blog, 2026-09 Power BI Blog, 2026-07 Fabric IQ overview Fabric IQ MCP server: getting started Microsoft Learn Microsoft Learn Skills for Fabric (agent skills) What's new in Fabric Apps GitHub (Microsoft) Fabric Blog, 2026-09 Direct Lake and Import composite models Strong/weak relationships in Power BI Power BI Blog, 2026-02 SQLBI From Dataflow Gen1 to Gen2 Azure Databricks architecture whitepaper Power BI Blog, 2026-04 Power BI Blog, 2026-09 September 2026 Feature Summary Microsoft Fabric Roadmap Power BI Blog, 2026-09 / Fabric version roadmap.fabric.microsoft.com Linked pages (accessed 2026-10-04) Q&A Send us your questions QR code 65 # Page. 66 ![Page Image](https://bcdn.docswell.com/page/GJWG4XGM72.jpg) Thank you Questions & feedback: Contact the presenter of this deck Latest version: Shared URL (added when distributed) QR code Please take our survey