The industry-specific problem
Bank transfers, card settlements and supplier invoices often use different descriptions. One bank debit may settle several invoices, while a payout may be net of fees. A text similarity score can conceal the mismatch.
Step-by-step workflow
- Retrieve permitted Hubdoc documents and QuickBooks Online ledger records in n8n; verify your extraction and export access.
- Normalize currencies and decimal amounts with accounting-safe rules and preserve original source IDs.
- Use a supported Claude model for candidate account or document matches, replacing the retired Claude 3.5 Sonnet API dependency.
- Apply deterministic amount/date checks and evaluate any proposed confidence threshold against labeled historical examples; 95% is not meaningful without calibration.
- Create a review packet for split payments, fees, tax differences and unmatched items. Post only through approved QuickBooks operations with accountant authorization.
Visual workflow
Hubdoc evidence + QBO ledger → n8n amount/currency normalization
→ Claude candidates → exact-match + calibrated policy
ambiguity → accountant queue
approved → QBO posting/reconciliation step
Specific configuration example
This n8n Code-node fragment consumes a normalized, authenticated event from preceding integration steps. It demonstrates the decision boundary, not a complete API client. Fields are workflow-local mappings, not claimed vendor field names. Upstream credentials, schema validation, permissions, durable reservations and provider operations must be implemented separately. Numeric policy examples are illustrative, not Imperio service terms.
// n8n Code node: exact amount is necessary, not sufficient
const match = $json;
const exact = match.bank_currency === match.document_currency &&
Number.isInteger(match.bank_amount_minor) &&
match.bank_amount_minor === match.document_amount_minor;
return [{ json: { bankTxnId: match.bank_txn_id,
documentId: match.hubdoc_document_id,
candidateAccepted: exact && match.date_window_valid &&
match.tax_reviewed && match.accountant_approved,
exception: match.split_payment || match.net_of_fees || !exact,
modelConfidenceIsApproval: false } }];
Outcome to measure
Measure accountant review minutes, unmatched transactions and post-close corrections. Illustratively, three minutes saved on 200 ordinary matches releases ten hours; include exception handling and correction costs in the evaluation.
Compared with manual work or other tools
Native bank rules are preferable for predictable recurring transactions. Claude assists with messy descriptions; deterministic reconciliation and accountant judgment remain responsible for postings.
Work with Imperio Automation
Ask Imperio to build a source-backed QuickBooks reconciliation queue using your approved Hubdoc records.
Frequently asked questions
Can a 95% model confidence score auto-post entries?
Not safely without validated calibration and independent accounting controls.
What about one payment covering several invoices?
Use a split-match review with the summed amounts and supporting documents.
Can it infer tax from a supplier name?
No. Tax treatment requires the transaction facts and authorized accounting rules.