Finance Automation for Dynamics 365 Business Central

Most "Business Central integration" pages describe a connector that could, in principle, be configured. This one describes two systems that run in production against Dynamics 365 Business Central right now: a collections workspace managing $11.8M of open receivables across 7,700+ invoices for a US insurance services firm, and an AI collections agent that drafts every chase email for a European equipment rental group's €7.6M ledger.

Both follow the same principle: Business Central stays the system of record and stays read-only. The automation layer is a custom platform in the client's own Azure tenant that reads BC data, does the work BC was never designed to do — chasing, scoring, promising, escalating — and stores its own activity data in its own database. Nothing writes back to the ledger.

We don't screen-scrape and we don't install anything inside BC. The data path is the one Microsoft built for exactly this:

• Azure Synapse or Microsoft Fabric mirrors BC data on a schedule — customers, invoices, payments, ledger entries — into a read-only analytical layer. The platform queries that mirror, never the live ERP.
• Microsoft Graph sends email from the client's own mailboxes, so every chase email comes from a real address the customer already knows.
• Microsoft Entra ID handles sign-on. Finance staff use the account they already have; access is role-based and revocable by the client's own IT.
• Azure PostgreSQL holds everything the platform creates — tasks, call logs, promises to pay, approval decisions. The ERP's books and the platform's activity trail stay cleanly separate, which auditors like.

If you already run Synapse or Fabric, most of the plumbing exists. If you don't, standing up the mirror is part of the build, not a prerequisite project.

The client had no system of record for collections — the work lived in personal inboxes next to a perfectly good ERP. We built a workspace that mirrors Business Central daily through Azure Synapse, re-scores every payer's risk each morning with a balance-weighted algorithm the finance team can explain to an auditor, generates and routes follow-up tasks across four escalation tiers, tracks promises to pay, and sends dunning email through Microsoft Graph. The ledger data it started from had inconsistent status flags and free-text payer names; the build absorbed that with resolution logic and review queues rather than demanding a cleanup project first.

Read the full case study, the guide to AR automation on Business Central, or the insurance services industry page this build anchors.

A Norwegian equipment rental group (~€30M revenue) ran a €7.6M receivables ledger with a one-person AR team. Their BC data already landed in a Microsoft Fabric Gold layer; we built an agent that reads it every morning, drafts the correct chase email for each account in Norwegian, classifies every customer reply into one of twelve intents, and routes everything through a human approval queue — no email leaves without sign-off. A single agent run drafted 162 correctly-staged chase emails. Before it touched a real customer, the agent passed a 23-scenario validation harness covering every outbound stage and inbound intent.

Read the full case study, the Microsoft Fabric integration page, or the equipment rental industry page this build anchors.

Business Central keeps what it's good at: the ledger, posting, credit memos, the audit trail of record. The platform owns what BC has no model for: who was chased and when, what was promised on which call, which dispute is blocking a six-figure payment, which payer's risk score just jumped. That split is deliberate. There is zero risk of automation corrupting the books, and if you ever re-implement or switch ERPs, the workflow layer — and its history — survives.

Everything a finance team needs to tune lives in an admin screen, not in code: templates, cadences, risk thresholds, escalation rules, even the AI prompts. No change request to us for a wording tweak.

Sometimes an off-the-shelf AR or AP app is the right answer, and we'll say so in a free assessment. The cases that come to us are the ones where it isn't: a payer structure no template models (one client chases ~2,500 insurance claims adjusters spread across 1,200 carrier customers), a security review that a third-party SaaS holding a copy of your ledger won't pass, or per-user-per-year pricing that inverts against a one-time build the moment the team grows. A custom platform on infrastructure you already pay for typically runs on tens of euros a month of Azure spend — and it fits the process you actually run, not the one the template assumes.

If your receivables or payables process runs on spreadsheets and inboxes next to Business Central, start with the AI readiness assessment — eight questions, a tailored reply within one business day. We'll map what an automation layer on your BC environment would look like, whether Synapse or Fabric is the right mirror for your setup, and what the first build should be. Or tell us how the process works today and we'll take it from there.