Integrations

Finance Automation on Microsoft Fabric

Most companies build a Fabric medallion architecture, point Power BI at the Gold layer, and stop. We build systems that read Gold and act — a production AI collections agent runs on a client's Fabric deployment today, drafting every chase email from Gold-layer invoice data with human approval on everything.

GDPR CompliantIrish CompanyXero & QuickBooks Integration

Key Takeaway

Most companies build a Fabric medallion architecture, point Power BI at the Gold layer, and stop. We build systems that read Gold and act — a production AI collections agent runs on a client's Fabric deployment today, drafting every chase email from Gold-layer invoice data with human approval on everything.

The Step Most Fabric Deployments Stop Short Of

The standard Microsoft Fabric story ends at a dashboard: land ERP data in the lakehouse, refine it through Bronze, Silver and Gold, connect Power BI, done. That's reporting — someone still has to read the dashboard, decide what to do, and go do it in an inbox.

The Gold layer is also a clean, consolidated, already-governed source for systems that do the work themselves. Every quality problem that would sink an automation project — duplicate customers, unreconciled payments, inconsistent codes — has already been solved in the medallion refinement your data team built. Automation gets that for free.

The Production Build: An AI Collections Agent Reading Gold

A Norwegian equipment rental group (~€30M revenue, Dynamics 365 Business Central, an existing Fabric deployment) ran a €7.6M receivables ledger with a one-person AR team and 1,080 open invoices. We built an agent that:

  • reads open invoices from the Fabric Gold SQL endpoint every morning,
  • works out the correct chase stage for every account and drafts the email — in Norwegian — grounded strictly in ledger data,
  • classifies every customer reply into one of twelve intents (promise to pay, dispute, invoice copy request, bankruptcy mention…),
  • and routes every draft through a human approval queue. Nothing sends without sign-off.

A single agent run drafted 162 correctly-staged chase emails — a workload that had previously been one person's week. Before go-live the agent passed a 23-scenario validation harness covering every outbound stage and inbound intent.

Read the full case study, or the deeper guide to Microsoft Fabric for finance teams.

The Architecture: Fabric Stays Read-Only

The agent treats Fabric exactly as it treats the ERP behind it: a read-only source. It queries the Gold SQL endpoint; it writes nothing back to the lakehouse and nothing to Business Central. Its own state — drafts, approvals, reply classifications, chase history — lives in an Azure PostgreSQL database in the client's tenant. Email goes out through Microsoft Graph from the client's own mailboxes; sign-on is Microsoft Entra ID; scheduled runs are Azure Container Apps jobs; managed identities mean no connection-string secrets to rotate.

The result is a clean separation your auditors and your data team will both accept: the medallion pipeline keeps its single responsibility, the ERP keeps the books, and the platform keeps the activity trail.

Streamline Your Financial Operations

Join hundreds of businesses already saving time with FinTask. Get a personalised demo today.

Do You Need Fabric for This?

No — Fabric is one good mirror among several. The same collections pattern runs for another client against Business Central through Azure Synapse, and the job-costing pattern runs against Sage through a REST layer. If you've already invested in Fabric, though, you're closer to operational automation than you probably think: the expensive part — clean, consolidated finance data — is done, and the missing piece is the workflow layer that reads it and acts.

If you're weighing Fabric up, do the two reporting basics first (consolidated AR and AP views in Gold, refreshed daily), then ask what the first system to act on those views should be. That's usually collections.

Where to Start

If Fabric is live and finance work still happens in inboxes and spreadsheets, the gap is a build, not a licence. Start with the AI readiness assessment — we'll map what already lands in your Gold layer, what state it's in, and what the highest-leverage system on top of it would be. Or tell us about your Fabric setup directly.

Frequently Asked Questions

Does the automation write anything into Fabric?

No. The platform reads the Gold SQL endpoint and nothing else — it writes no tables, modifies no pipelines and never touches the ERP behind Fabric. Its own operational data lives in a separate Azure PostgreSQL database in your tenant, so the medallion architecture keeps its single responsibility.

We only use Fabric for Power BI. Is that enough to build on?

Usually, yes — if your Gold layer already carries customers, invoices and payments for reporting, an operational system can read the same tables. The assessment establishes whether the Gold model needs extending (for example, adding payment application detail) before the first build.

Can the AI send emails without human review?

Not in our builds. Every customer-facing draft goes through an approval queue, with hard stops for legal threats and bankruptcy mentions in replies. In production this costs the approver minutes a day and is precisely what makes a one-person AR team comfortable running an agent against a €7.6M ledger.

We're on Azure Synapse, not Fabric. Does the pattern apply?

Directly. One production collections build runs on Business Central mirrored through Synapse rather than Fabric — the workflow layer is identical, only the read endpoint differs. If you're mid-migration from Synapse to Fabric, the platform moves with you by swapping the connection, not the build.

Reza Shahrokhi, ACA - Chartered Accountant and FinTask Founder

Written by Reza Shahrokhi ACA

Chartered Accountant (Chartered Accountants Ireland) • Founder of FinTask • 8+ years in finance & automation

Reza is a Chartered Accountant and the founder of FinTask. He specialises in helping growing businesses automate accounts payable, invoice processing, and financial reconciliation using AI-powered tools integrated with Xero and QuickBooks.

More about the author

Fabric Is Live. Now Make It Do the Work.

Tell us what lands in your Gold layer and we'll map the first system that should read it and act — with every AI action behind a human approval queue.