Intercompany reconciliation with AI: match counterparties, FX and timing breaks

When to use it

Use it for multi-entity groups where intercompany doesn't net to zero at consolidation and someone spends days chasing the breaks over email. It's most valuable when entities are on different ERPs or currencies, and where references never line up exactly.

Input data

• Intercompany balance by counterparty pair from each entity (receivables, payables, loans, and intercompany revenue/expense if applicable).
• Transaction detail for each intercompany account for the period (date, reference, description, amount, currency).
• Month-end and average FX rates from your approved rate source (don't let the AI look rates up).
• Your intercompany policy: settlement terms, which entity records first, and a rounding threshold.

The prompt

Replace the [bracketed] items. Paste it with your files attached. It works in any capable assistant; tool-specific notes are in the steps.

Help me reconcile intercompany balances for [period] across these entities: [list with functional currencies].

Attached:
- Intercompany balances by counterparty from each entity: [files]
- Transaction detail for intercompany accounts: [files]
- FX rates (month-end and average) from [source]: [file]
- Rounding threshold: [currency] [amount]

1. Build the counterparty matrix: for each pair (A↔B), A's balance with B and B's balance with A, both translated to [reporting currency] at the month-end rate, and the difference.
2. For each pair with a difference above the threshold, match the transaction detail between the two sides: exact reference, then amount within the rate tolerance, then fuzzy description/date (±[5] days). List the matched pairs.
3. Classify each remaining difference as:
   - Timing (recorded by one side only, with the date suggesting the other side will record next period)
   - FX (difference explained by rate used or translation)
   - Rounding (under threshold)
   - True break (missing, duplicated, wrong amount, wrong counterparty or wrong account)
4. For true breaks, propose which entity should correct and draft the entry (Dr/Cr, account, amount, memo), labeled DRAFT – for review, with the evidence.
5. Summarize: pairs reconciled, total difference by category, and the list of items to send to each entity's accountant.

Use only the rates provided. Don't force pairs to agree.

Steps

1. Collect balances and detail from each entity after their sub-close, at a common cut-off.
2. Run the prompt and confirm the matrix totals agree to each entity's trial balance.
3. Send each entity its list of breaks and timing items. The entity accountant confirms or explains.
4. Correcting entries are booked by the responsible entity through its normal JE approval.
5. Re-run after the corrections. Eliminations at consolidation should then net to within threshold.

Control map

What the AI may do, step by step. How ratings work.

Audit trail

• Keep the matrix, the matching output and each entity's confirmation of their breaks.
• Reference correcting JE numbers against each break on the final reconciliation.
• Document the FX rate source used.

Failure modes

• FX rates the model "knows" instead of your approved rates. Always provide the rates.
• Mis-signed balances. Receivables positive on one side, payables negative on the other. Confirm the sign convention in the matrix.
• Mismatched counterparty codes across ERPs ("UK01" vs "Acme Ltd"). Provide a mapping table.
• Labeling true breaks as timing because the date is near period-end. Check that the item actually appears next period.

Data handling

• Use a company-approved AI workspace whose terms say your inputs are not used for model training. Check your plan; don't assume it.
• Strip or mask what the task doesn't need: employee names, bank account numbers, customer PII, SSNs or TINs.
• If you are a public company, treat pre-release results as material non-public information and keep them inside approved systems.