A company's internal ledger and its actual bank statement should show the same cash position, but rarely match perfectly at any given moment: checks that haven't cleared yet, bank fees not yet recorded internally, deposits in transit. Reconciliation is the process of accounting for every one of these differences until the two balances tie out.

Most differences are routine and expected (timing lags, small recurring fees), but a genuine mismatch, a transaction on the bank statement with no internal record, or vice versa, needs investigation, since it can signal an error, a missed entry, or in rarer cases fraud.

Frequently Asked Questions

How often does bank reconciliation typically happen?

Monthly at minimum as part of the standard close process, though companies with high transaction volume or tighter cash management needs often reconcile daily or weekly to catch discrepancies faster.

What's the most common cause of a reconciliation mismatch?

Timing differences, a check written but not yet cashed, a deposit made but not yet processed by the bank, account for the large majority of mismatches, with genuine errors or fraud being comparatively rare but the reason every mismatch still gets checked.

Why is reconciliation a common first process for AI automation?

It's high-volume, rules-based for the majority of transactions, and has a clear, checkable outcome (the balances either tie out or they don't), making it easier to validate an AI agent's accuracy against than more subjective work.

Can reconciliation be fully automated end to end?

The routine matching can be, largely. Genuinely unexplained discrepancies, ones that don't fit a known pattern like a recurring fee or timing lag, typically still get escalated to a person, since they may indicate something that needs real investigation rather than a mechanical match.