An intercompany reconciliation is the check that two companies in the same group record their dealings with each other identically: what one shows as a receivable from its sister company, the other shows as an equal payable. The balances must agree so they cancel out on consolidation. Every difference is explained and one side, by agreed rule, corrects.
Why does an intercompany reconciliation matter once agents act?
Intercompany matching across time zones, currencies and systems is repetitive and email-heavy, which makes it an obvious job for an agent. It is also a job where an agent can quietly make things worse, because both records belong to the group and neither is an independent source.
In a bank reconciliation the bank's statement is the arbiter. In an intercompany reconciliation, there is no arbiter unless the group writes one down. An agent asked to "fix the difference" will pick a side. The group's policy, not the agent, should decide which side adjusts. That rule is the kind of thing AI agents for finance need to be given explicitly.
How does it work in practice?
- Pull intercompany balances and transactions for each pair of entities at the same cut-off.
- Translate to a common currency at the group's agreed rate.
- Match transactions: recharges, loans, interest, dividends, trading invoices.
- Classify differences: timing (one side booked, the other not yet), exchange rate, wrong counterparty code, or a genuine disagreement about the amount.
- Apply the group rule on which side adjusts, usually the side that issued the charge.
- Both entities' controllers agree the result before consolidation.
An agent can do steps 1 to 4 and draft the adjusting journals in step 5. Step 6 stays with people in each entity, because each controller answers for their own books. The detailed process is in how to prepare intercompany reconciliations with an AI agent, and the limits are discussed in can an AI agent do intercompany reconciliations?
What does that look like in an example?
A group's UK entity recharges 50,000 of shared IT costs to its German entity on the last day of the month. The UK side books a receivable of 50,000. The German side books the invoice three days later, in the next period, and at a different exchange rate. At month end the pair shows a difference of the full amount, and the following month a small exchange difference remains.
The agent classifies the first as timing and the second as rate, and drafts an accrual in the German books for the month-end difference, citing the UK invoice. The German controller approves it. The rate difference is booked according to the group's rate policy, which the agent quotes in its note.
How is it different from a bank reconciliation?
A bank reconciliation compares your books with a record you do not control, so the bank's figure usually wins and differences are mostly your errors or timing. An intercompany reconciliation compares two records you do control. Either side could be wrong, both might be, and the difference must be agreed rather than simply corrected. The general principles of reconciliation apply to both, but intercompany adds a negotiation and a rule for who gives way.
Which related terms should you know?
- Elimination: removing intercompany balances and transactions on consolidation so the group does not count trade with itself.
- Transfer pricing: the rules for how group entities charge each other, which an agent should read but not set.
- A bank reconciliation template is a useful starting layout for intercompany schedules too.
Questions people ask
Who should approve an intercompany adjustment drafted by an agent?
The controller of the entity whose books change. If both entities adjust, both controllers approve their own side. Group finance can set the rule and resolve disputes, but should not post into an entity's ledger on that entity's behalf.
Why do intercompany differences keep recurring?
Usually because of a structural cause: mismatched cut-off dates, different exchange rate sources, or a counterparty code that one entity never uses. An agent that classifies differences month after month can show which pair and which cause recur, so the fix happens at source rather than at every close.