Reconciliation still needs a finance owner to decide what a difference means. Strawberry can collect the relevant statements, ledgers, invoices, and browser-based account records, then prepare a dated exception list with the evidence beside each item.
01
Define what counts as a match before asking AI to find exceptions.
An amount alone is rarely enough.
Dates move, fees are netted, descriptions vary, and one payout can represent several orders.
Set the matching rules first: permitted date window, currency treatment, reference priority, rounding threshold, and who owns ambiguous items. Strawberry can apply those rules to produce a result that a reviewer can trace.
02
Let the evidence travel with the discrepancy.
A reconciliation issue becomes expensive when the reviewer has to reopen six systems just to understand it. The exception should carry the reference, amount, dates, links, and the reason it was flagged.
Strawberry can assemble that case file from the sources you provide. The output is a compact review pack, not a black-box “matched” label.
03
Separate routine timing differences from genuine errors.
A charge posted one day late is different from a duplicate payment.
A good check distinguishes timing, missing documents, duplicate entries, and amounts that need investigation.
Ask Strawberry to classify the exception without deciding the accounting treatment. That gives the finance team a cleaner queue and preserves the judgment required for the close.
04
Make the monthly close a repeatable review.
At every close, the reconciliation pass can apply the approved account-matching rules, collect the supporting entries, and isolate unmatched or duplicate transactions for review. It must not post an adjustment or clear an exception on its own: an unreviewed match can hide an unsettled balance, duplicate payment, or timing error.
The routine can prepare the queue; finance decides which entries are corrected and why. That distinction matters when a number flows into management reporting or a statutory record.
It can compare statement lines with the ledger and supporting records using the matching rules you define, then surface exceptions for review.
Do not make that the default. Keep financial postings approval-gated and use the workflow to prepare the supporting evidence first.
Define the matching policy for partials and grouped payouts, then require the output to show every linked transaction.
At minimum: amount, dates, reference, likely match candidates, source links, and a clear reason for the flag.
Yes. At each month-end close, rerun the approved matching rules and prepare an exception pack with supporting entries, unmatched balances, and duplicates. Review proposed clears or adjustments before posting; a bad match can make the ledger look reconciled while cash or liabilities remain wrong.