Day three of the close, and consolidation will not balance. The US parent shows a receivable from its UK subsidiary. The UK books show a smaller payable, in sterling, translated at a different rate. Part of the gap is a management fee the parent invoiced on the last day of the month, which the UK team has not yet seen. Part of it nobody can explain.
Intercompany reconciliation is simple in principle, because both sides of every balance sit inside the same group and should cancel on consolidation. In practice it is one of the most common reasons a close slips, and the causes are structural rather than careless.
Why intercompany reconciliation breaks
Timing. The seller books the charge when it invoices. The buyer books it when the invoice arrives or is approved. A charge raised at month end often sits on one side only.
Foreign exchange. Each entity books in its own currency, at its own rate, on its own date. A correctly booked pair will not agree in group currency unless both sides use the same rate source and date convention.
Different charts of accounts. One entity books a management fee to revenue, the other to a cost line with no intercompany flag. Matching then starts with a hunt for where the other side went.
Different ERPs. The parent may run SAP or Dynamics 365 while an acquired subsidiary runs QuickBooks Online or a local package. The trading partner field one system enforces may not exist in the other, and references get truncated in exports.
One-sided postings. An accrual, a recharge or a correction posted by one entity without notice to the other. These turn a timing difference into an unexplained one.
The problem is widely reported. In research commissioned by BlackLine, a finance software vendor, and run by Dimensional Research (published June 2023; 263 people responsible for intercompany at international companies with revenue above $500 million), "99% said they face specific challenges with intercompany financial processes". Nearly all (97%) had had to resolve multi-million-dollar material variances. It is vendor research from large groups, but smaller groups will recognize the pattern.
Matching rules that hold up
Matching works best as rules applied in order, from strict to loose. Each tier pairs what it can and passes the rest down.
| Tier | Matches on | What usually falls through |
|---|---|---|
| 1. Exact | Shared reference, counterparty code and amount in transaction currency | Missing or reformatted references |
| 2. Reference with tolerance | Shared reference and counterparty, amount within an agreed FX tolerance | Bank charges deducted at settlement |
| 3. Counterparty and amount | Counterparty code, transaction-currency amount, dates within a short window | Bundled charges, one-sided accruals |
| 4. Many-to-one | One settlement against several items on the other side | Disputed items, wrong counterparty |
| 5. Exception | Everything left | Goes to a named owner with a due date |
Three design choices matter more than the tiers. Match in transaction currency, so exchange rates do not create false differences. Make the counterparty code mandatory on every intercompany posting in every system, even where that means a manual field in a smaller ERP. And have group finance write the tolerances down, so they do not drift with whoever runs the close.
A netting and settlement routine
Matching tells you what disagrees. A settlement routine stops balances building up. A workable monthly cycle has six steps.
Cut-off. Intercompany charges are raised by a fixed working day before month end. Later charges move to next month, on both sides.
Seller books, buyer mirrors. The entity raising the charge books it and sends the document. The receiver books the mirror entry with the same reference and transaction-currency amount, not its own estimate.
Confirmation. Each entity confirms balances with every counterparty and raises differences by a set date, with the document.
Dispute deadline. Disputes not raised in time are booked as the seller recorded them and resolved next month.
Netting statement. Group treasury issues a statement of what each entity pays or receives, net.
Settlement. Both sides post the settlement with the netting reference on the same date, with realised FX to the same accounts.
Some countries restrict cross-border netting or attach approvals or withholding to intercompany payments. Confirm each entity's position with local advisers before including it.
Where AI helps with matching
Rules handle the clean items. AI is useful in the layer below, where a person currently reads descriptions and makes judgment calls.
Suggesting matches. A model can read free-text descriptions, partial references and amounts that differ by bank charges, and propose a pairing with a stated reason. Microsoft's prerelease documentation for its Financial Reconciliation agent in Excel describes a report that sorts transactions into "unmatched, potentially matched, and perfectly matched". The middle group is where a person still decides.
Explaining differences. A first-draft classification of each open item: timing, FX, missing posting, wrong counterparty or dispute.
Drafting the query. A message to the counterparty's accountant with the item and the document, for a person to check and send.
A person should accept any match that leads to a posting and decide which entity books a correction. Two failure modes are worth watching. Unrelated items with the same amount get paired with confidence, and a model trained on a history of forced matches repeats them. Log accepted and rejected suggestions and review a sample each month.
What AI matching will not fix
It will not fix missing data. If one system does not record a counterparty or a reference, the model is guessing, and a well-presented guess is still a guess. The counterparty code and booking rules come first.
It will not settle disagreements about substance. Whether a recharge should exist, at what price, and which entity bears a cost are questions for transfer pricing policy and management.
Frequently asked questions
What is intercompany reconciliation?
Intercompany reconciliation is the process of agreeing the balances and transactions that entities in the same group record with each other, so they eliminate cleanly on consolidation.
Why don't intercompany balances match?
They usually differ because of timing, exchange rates, different charts of accounts, different ERPs or postings made by one side alone. Most differences become explainable once both sides use a shared reference and a common booking rule.
Should intercompany matching use group currency or transaction currency?
Match in transaction currency. Group currency adds differences caused only by the rates each entity used, which hides real exceptions among false ones.
Can AI do intercompany matching on its own?
No, AI can suggest matches and explain differences, but a person should accept any match that leads to a posting. Models can pair unrelated items that share an amount.
Where CFOLogic fits
We design and build AI-assisted matching and exception views on top of the ERPs a group already runs, read-only by default, with a person approving any posting or message that leaves the team. We usually start with a proof of concept scoped to one intercompany flow. Our CFO Success Partners page explains how that work is organised.