Reconcile subledgers to the general ledger more often, so month-end starts from exceptions your accountant has already reviewed

Month-end often begins with a subledger that does not tie to the general ledger, and someone then spends days finding out why. Reconcile those subledgers every day instead, and your close starts from a short list of exceptions your accountant has already reviewed.

The method fits any ERP whose ledgers can be exported or queried: compare them daily, sort each difference into a named exception type, and let an agent propose a fix with evidence for your accountant to decide.

Continuous reconciliation is comparing each subledger with the general ledger on a short, regular cycle, so differences are explained while fresh.

Why reconcile more often instead of faster at month-end?

In Planful's 2025 Global Finance Survey, a vendor survey, 37% of finance teams said account reconciliations take more than six days, and 45% reported data reconciliation issues.

A daily cycle gives you:

• Less chasing. Each difference surfaces within a day of posting, while whoever posted the voucher still remembers it.
• A calmer close. Month-end inherits a reviewed queue instead of a backlog.
• More judgement. Software assembles evidence; your accountant decides.

Microsoft presents its Dynamics 365 Finance Account Reconciliation Agent, a "production ready preview" whose documentation was last updated 2026-09-08, as a step toward a continuous close. Its workspace reconciles the general ledger with the payables, receivables, tax and bank subledgers on a schedule you define, in the background or during off-hours, replacing SSRS reports.

Five operating rules that let you delegate daily reconciliation with control

Functional AI Solutions builds reconciliation workflows on five operating rules. Microsoft's documented agent illustrates each, and each can also be built around an older ERP:

1. Name a small exception catalogue. The agent processes two exception types: voucher amount mismatch, and pending accounting transferred to the general ledger.
2. Fix the permitted actions for each type. For a voucher amount mismatch, the agent can recommend Create journal entry; the user accepts it or chooses Reverse, Link transactions or Accept without change.
3. Log everything and make it reversible. Each exception records actions by users, automation or agents, and a reconciled item can be undone; the workspace asks for a reason when you undo.
4. Cap the volume and set alerts. You set daily or monthly exception limits, with optional alerts at 50%, 75% or 90% and a notice when a run is skipped at the limit.
5. Keep the decision with a named accountant. The agent does not run until someone confirms its settings, and a person submits each action, singly or for a group of similar exceptions.

The agent's documentation is prerelease, and Microsoft may update the AI model automatically. Its enhancements release plan lists general availability as Sep 2026, notes that some functionality is not yet released and aims at fully autonomous reconciliation, so decide now which decisions stay with your accountant.

Daylit's bank reconciliation skill also routes exceptions to a human review queue with reason codes, and Taiwan's Nocoo lets an AI agent prepare accounting drafts that a person posts and approves.

What does an agent need before it can explain a mismatch?

An agent explains an exception by following the links behind it:

Customer invoice → receipt → voucher → subledger posting → general-ledger account → legal entity → responsible person

A vague account mapping or missing link leaves the exception unexplained. Microsoft's workspace, too, takes its account list from posting profiles and the bank and tax account setup.

A business ontology is a shared map of your business objects, such as customers, invoices, vouchers, accounts, entities and owners, and of how they relate. FAIS builds that foundational ontology over the systems you already run, with a connector for each, so the agent receives the invoice, receipt, voucher and owner behind a difference, not a bare variance. The same map then serves your next finance workflow without starting over.

Before and after: an illustrative Taiwan close

Consider a Taiwan components manufacturer with two legal entities, an ageing ERP and receivables tied out in Excel.

Before: On the first day of close, the receivables subledger and its control account disagree. Two weeks earlier, someone booked a customer rebate as a manual journal straight to the control account. The accountant exports both sides, sorts vouchers, asks sales and corrects it days later.

After: The next morning's run flags the difference as "in ledger, not in subledger", and the ontology attaches the voucher, its poster and the rebate's customer. The agent proposes reversing the journal and issuing a credit note in receivables. The accountant confirms with sales and accepts; the log records the decision. The controller then blocks manual entry on the control account, a fix of the kind Microsoft's release plan describes Copilot proposing.

How should you pilot daily reconciliation?

Start with receivables at one entity, daily, for one close. FAIS can build the workflow on your existing ERP and spreadsheets, or operate a reconciliation agent you already run and tune its limits and escalation rules. Measure against your last close:

• open exceptions at period end
• time to clear each exception
• recommendations accepted versus changed
• undos and reversals
• runs skipped because a limit was reached

Which reconciliation costs your close the most days? Talk it through with us: its exception types, the actions you would permit and the links an agent needs to explain each difference.