Current as of 16 October 2026. Microsoft licensing, prices and export options change often; we update this post when they do.
The Microsoft data stack for Dynamics 365 has five layers: the D365 ledger, a route to get data out, a place to store and reconcile it, a Power BI semantic model that defines the numbers, and the reports and Copilot on top. Finance can only trust the top layer if every layer underneath ties back to the ledger.
Microsoft sells the pieces separately, so most finance leaders see product names rather than a picture. Here is the stack as finance experiences it, from the ledger up.
| Layer | Microsoft product | What it does | What finance gets | If it is weak |
|---|---|---|---|---|
| 1. Ledger | Dynamics 365 Finance and Operations or Business Central | Records every transaction and posts the books | The figures you sign off and the auditor tests | Nothing above it can be right |
| 2. Getting data out | Link to Fabric or Azure Synapse Link (F&O); Power BI apps, APIs or mirroring (BC) | Copies ledger data to somewhere it can be reported on | Reporting that keeps working when Microsoft changes the feeds | Board packs stop updating, or update with gaps |
| 3. Storage and reconciliation | OneLake, Fabric Lakehouse or Data Warehouse | Holds the data and checks it agrees with the ledger | One group figure across entities that ties back to D365 | Entities that do not add up, and a manual tie-out every month |
| 4. Meaning | Power BI semantic model | Defines revenue, margin, cost centres and hierarchies once | One version of the numbers for finance, the board and budget holders | Every report calculates margin its own way |
| 5. Insight | Power BI reports, Excel, Copilot | Shows the numbers and answers questions about them | Faster answers and a board pack people trust | Polished visuals, or confident AI answers, built on wrong figures |
The point of the table is the last column. Most reporting problems finance teams describe as "Power BI issues" actually start at layer 2 or 3.
D365 is built to run transactions, not to answer board questions across entities, years and dimensions. Its standard reports work well inside one company. They struggle when the CFO wants a consolidated view with eliminations, a five-year trend, or a figure that combines ledger data with budgets held elsewhere.
So the data gets exported to Excel, and the close starts to depend on whoever owns the workbook. We covered the warning signs in Power BI vs Excel for month-end. Dynamics 365 reporting at group level needs the layers above the ledger, built so they reconcile back to it.
This is the layer most likely to change under you, and the route differs by product. Your IT team will want the detail in our Fabric Link vs Synapse Link guide. Here is what finance needs to know.
For F&O, Microsoft supports two Dataverse routes. Link to Microsoft Fabric makes D365 tables available in Fabric without building a copy process, and creates a Lakehouse, a SQL endpoint and a Power BI semantic model automatically. It needs Power BI Premium or Fabric capacity in the same region as your Dataverse environment, and it uses Dataverse storage, so it is not free. Updates can take up to 60 minutes, with a faster sync rolling out.
Azure Synapse Link for Dataverse exports to your own storage, with incremental updates set anywhere from 5 minutes to 24 hours. From 15 October 2026, its Delta/Parquet option is closed to new customers, so new setups now lean towards Link to Fabric. For finance, the question is simple: how fresh do the numbers need to be at close, and who pays for the storage?
Business Central works differently. Export to Data Lake and Synapse Link do not apply. Microsoft's Power BI apps for Business Central (Finance, Sales, Purchasing, Inventory and others) are free, but each user needs a paid Power BI licence and the apps work per company. The Finance app only shows correct KPIs once you map your G/L account categories, which is a finance task.
For anything beyond the apps, Microsoft's extraction guidance points to APIs and data integration tools. Mirroring Business Central to Fabric is coming: Microsoft plans a production-ready preview in October 2026, timing may change, and it will need BC version 29.0 and Fabric capacity. Plan for it, but do not build the close on it yet. Microsoft's open-source bc2adls tool is an experiment, not a supported product.
Microsoft's transition FAQ confirms Export to Data Lake stops permanently on 30 November 2026, with no extensions. Files already exported stay where they are, but no new or changed data arrives after that date. If a board pack, forecast or bank covenant report still reads from it, the figures quietly freeze. BYOD has no retirement date, though Microsoft recommends moving to Synapse Link or Link to Fabric.
OneLake is a single, tenant-wide data lake inside Fabric. It is a good place to keep D365 data. Storing data is not the same as reconciling it, though.
A Fabric Data Warehouse is a relational warehouse built on that lake. The warehouse is where finance logic lives: intercompany eliminations, currency translation, mapping charts of accounts across entities, and the checks that prove each balance agrees with the ledger. That is what gives the group one figure that ties back to D365, and it is what auditors and the board ask about.
For a single Business Central company with standard reporting, OneLake and the Power BI apps may be enough. For multiple entities, mixed charts of accounts or eliminations, a reconciled warehouse layer is usually the practical answer. Our business guide to data warehouses covers the reasoning.
Microsoft Fabric is Microsoft's analytics platform: OneLake, data engineering, warehousing, Power BI and Copilot on one platform. Power BI is part of Fabric, but you can still use Power BI with Pro or Premium Per User licences and no Fabric capacity at all.
You need Fabric capacity when you want Link to Fabric for F&O, a Fabric warehouse, Copilot in Power BI, or many report viewers without individual licences. On F64 or larger, viewers with a free licence can read reports; below F64, every viewer needs Pro or Premium Per User.
At the time of writing (16 October 2026), Microsoft's Fabric pricing for UK South on pay-as-you-go works out at about GBP 231 a month for F2 and about GBP 7,405 a month for F64, running all month. Power BI Pro is GBP 10.80 per user per month and Premium Per User GBP 18.50, excluding VAT and paid annually.
One cost point for finance: capacity overage is switched on by default for new Fabric capacities. If usage spikes, Microsoft bills the excess at three times the pay-as-you-go rate, and the threshold you set is not a hard cap. Ask who sets that threshold and who sees the bill before the capacity goes live.
The reports are the visible part. The part that matters is the semantic model, which Microsoft describes as a source of data "ready for reporting and visualization". It is where revenue, gross margin, cost centre hierarchies and period logic are defined once.
When the semantic model is right, every report, every budget holder and every Excel PivotTable reads the same definitions. Finance can keep working in Excel through Analyze in Excel, connected to governed numbers rather than last month's export. When the model is wrong, or there are five competing models, you get five versions of margin and a board meeting spent arguing about which is right. Our Power BI for Dynamics 365 service starts from the model, not the dashboard.
Copilot sits at the top of the stack and reads from everything below it. Copilot in Power BI needs paid Fabric capacity (F2 or above) or Power BI Premium (P1 or above); Pro or Premium Per User licences alone are not enough. For capacities in the UK, it stays off by default until an admin allows data to be processed outside the UK, which is a governance decision finance should be part of.
Microsoft is adding finance-specific AI too. The Financial Reconciliation agent in Excel is generally available, and the Account Reconciliation Agent for D365 Finance is still in production-ready preview. Fabric IQ in Microsoft 365 Copilot answers questions from governed Power BI models. Microsoft Ignite runs from 17 to 20 November 2026, so expect more announcements; we will update this post.
Because Copilot does not know your numbers are wrong. Microsoft's own guidance, Prep data for AI, warns that without preparation, outputs can be "generic, inaccurate, or even misleading". A dashboard with a wrong figure at least gets questioned. A fluent answer from Copilot often does not.
Preparing the semantic model helps, but it cannot fix data that never reconciled to the ledger in the first place. The order matters: ledger, a stable route out, reconciliation, one semantic model, then AI. Skip the middle and Copilot gives the CFO a confident answer built on the wrong group figure.
That middle layer is what Reveal provides. It is our pre-built, self-reconciling data warehouse foundation for Dynamics 365, so the checks against the ledger are part of the foundation rather than a project phase you build and maintain yourself. If you would rather not run the stack in-house, our managed BI for Dynamics 365 service runs it for you. For short answers to common questions, see our answers hub.
No. Power BI works with Pro or Premium Per User licences and no Fabric capacity. You need Fabric capacity for Link to Fabric from Finance and Operations, a Fabric warehouse, Copilot in Power BI, or letting free-licence users view reports on F64 or larger.
No. Power BI is licensed separately from Dynamics 365. Business Central includes free Power BI apps, but each user still needs a paid Power BI licence, and Finance and Operations reporting through Fabric needs Power BI Premium or Fabric capacity.
No new or changed data flows after 30 November 2026, and Microsoft has said there will be no extensions. Files already exported stay in place, so any report reading from them shows frozen figures. Microsoft's supported replacements are Azure Synapse Link for Dataverse and Link to Microsoft Fabric.
Often, yes. OneLake stores data, but it does not reconcile it. If you have multiple entities, eliminations or different charts of accounts, a warehouse layer that checks every balance against the D365 ledger is usually the practical way to get one group figure you can trust.
No. Microsoft's own guidance says unprepared data can produce generic, inaccurate or even misleading outputs, and preparation cannot fix numbers that never matched the ledger. Reconcile the data to D365 first, define it once in a semantic model, then point Copilot at it.
If you are planning Fabric, Power BI or Copilot on Dynamics 365, start with the layer that makes the numbers trustworthy. See how our Power BI for Dynamics 365 service builds reporting that ties back to the ledger, how Reveal gives you a reconciled foundation without a bespoke build, or talk to us about your stack.