How an AI agent can prepare a mismatch queue for accountant review
This workflow gives an AI agent a defined job, a bounded set of records, and a result a person can review. The agent reads the relevant Mercury context and matches it with Xero, applies the rules in the prompt, and keeps the source behind every recommendation. It returns a proposed handoff rather than taking consequential actions on its own.
Can an AI agent prepare a mismatch queue for accountant review?
Yes. Start with the scope, date range, decision rules, and fields that identify the right records. The agent can collect the evidence, compare states or sources, mark conflicts and missing data, and organize the result around the outcome above. A reviewer then checks the matches and judgment calls before approving messages, record updates, bookings, purchases, publishing, or other write actions. The guide below shows the records, boundaries, prompt, and handoff needed for this specific workflow.
Match before reviewing coding
Mercury supplies bank transactions, counterparties, dates, amounts, and status. Xero supplies ledger accounts, bills, invoices, bank-feed records, and reconciliation state. Together, they can prepare a month-end queue that shows exactly which bank movements do not have a clear accounting match.
Choose the entity, Mercury accounts, Xero organization, period, currencies, and matching tolerance. The agent should match on stable identifiers when available, then date and amount. Similar amounts alone can connect the wrong entries, especially for recurring subscriptions or batch settlements.
Example starter prompt
Compare Mercury transactions for [accounts] with Xero records for [organization] during [period].
Match with [transaction ID or approved rules]. For each result, show the Mercury transaction, Xero record, amount, date, currency, counterparty, ledger account, and reconciliation state. Separate exact matches, possible matches, bank-only items, and ledger-only items.
Do not create, edit, post, or reconcile entries. Return the month-end exception queue and the evidence needed for an accountant to decide each item.Keep accounting judgment with the accountant
The agent can point out missing descriptions, duplicate candidates, or coding that conflicts with supplied rules. It should not choose tax treatment, recognize revenue, or infer the business purpose of an unexplained transaction.
The final queue should preserve both source links or identifiers, the proposed next check, and the responsible owner. Corrections and reconciliation happen only after review in Xero.
Questions this workflow answers
Could an agent match bank activity to the ledger and give the accountant a month-end queue of missing, duplicate, and unexplained entries?
Yes. Mercury supplies posted and pending bank transactions, account, counterparty, amount, currency, date, and identifiers. Xero supplies ledger entries, bank-feed records, coding, tax fields, references, and reconciliation state. The agent joins records using identifiers where available, then amount, currency, date tolerance, and counterparty under rules the accountant approves.
Matches should be classified as exact, probable, one-to-many, many-to-one, or unmatched. Fees, batched deposits, split postings, transfers between accounts, and timing differences can make simple amount matching wrong. The agent should preserve candidate relationships and explain the evidence instead of forcing every bank line into a ledger entry.
The queue can identify missing records, duplicate candidates, inconsistent dates or amounts, absent descriptions, and coding that conflicts with supplied policy. It must not choose tax treatment, recognize revenue, or invent business purpose. Pending Mercury transactions should remain outside completed reconciliation unless the accounting process explicitly handles them.
The handoff includes both source identifiers, match class, discrepancy, supporting fields, proposed next check, and owner. An accountant reviews corrections and performs reconciliation in Xero. The agent reduces manual comparison while keeping ambiguous bundles and professional accounting judgments out of automatic writes.
Month-end matching needs to account for deposits that arrive net of fees and payouts that combine several underlying transactions. A $9,850 bank credit may correspond to a $10,000 receivable less a $150 fee, not to a missing $9,850 invoice. The agent should preserve the bundle and show the arithmetic, supporting references, and any unmatched remainder. Foreign-currency entries also need transaction currency, ledger currency, rate date, and recorded gain or loss kept distinct. Those cases go to the accountant with the candidate relationship, not a forced one-to-one match.