Reconcile Invoices and Bank Transactions Automatically
The AI Transaction Matcher runs a two-stage hybrid pipeline. Stage 1 narrows candidates with deterministic rules — currency, date window, and amount tolerance. Stage 2 applies AI-powered scoring that evaluates vendor names, descriptions, amounts, and dates. Matches above the confidence threshold are suggested with plain-English explanations. Confirmed matches auto-mark invoices as paid.
Key AI Transaction Matcher Capabilities
Built-in functionality that eliminates repetitive document tasks
Deterministic Prefilter (Stage 1)
Before the AI runs, an algorithmic filter narrows candidates using exact currency match, a date window of -7 to +60 days from invoice date, and an amount tolerance of plus or minus 10%.
AI Confidence Scoring (Stage 2)
Candidate pairs from Stage 1 are evaluated by an AI model that weighs vendor name similarity, transaction description relevance, amount closeness, and date proximity — returning a 0-to-1 confidence score per pair.
0.6 Confidence Threshold
Only matches scoring 0.6 or higher are presented as suggestions. This threshold is calibrated to deliver a 92% auto-match rate while keeping false positives to a minimum.
One-Click Confirm, Auto-Mark Paid
Confirming a match links the transaction to the invoice and automatically sets the invoice status to "paid." No manual status toggling — your records stay synchronized with your bank.
Plain-English Match Explanations
Every suggested match includes an AI-generated reason — referencing vendor name overlap, amount difference, and date gap — so reviewers understand the rationale without digging into raw data.
Manual Matching for Exceptions
Transactions the AI cannot resolve appear in the Unmatched tab. A manual matching interface lets you browse unmatched invoices and link them yourself, following the same confirmation workflow.
Inside the Matching Pipeline
AI identifies and extracts data from every supported format
Currency Gate
Only invoices and transactions in the same currency are paired. Cross-currency candidates are excluded before AI scoring begins.
Date Window (-7 to +60 Days)
Transactions are matched to invoices within a seven-day early window and a sixty-day late window — accommodating prepayments and typical net-30/net-60 terms.
Amount Tolerance (Plus or Minus 10%)
Transaction amounts must fall within 10% of the invoice total, accounting for partial payments, rounding differences, and bank fees.
AI Batch Evaluation
Candidate pairs passing the prefilter are sent to the AI model in batches of 20. The model scores each pair on vendor name, description, amount, and date signals.
Confidence Filtering
Pairs scoring below the 0.6 threshold are discarded. Only high-confidence matches reach the review queue — reducing noise for your team.
Confirmation and Payment Status
Confirmed matches create a reconciliation record and auto-update the invoice to "paid." Rejected matches are excluded from all future matching runs.
How It Works
From connection to first extracted invoice in under five minutes
Upload a Bank Statement
Upload in any of the 8 supported formats. Transactions are extracted by the Bank Statement Parser and become available for matching immediately.
Prefilter Narrows the Candidate Set
The algorithmic stage compares every extracted transaction against every unmatched invoice, applying currency, date, and amount rules to produce a shortlist.
AI Scores and Explains Each Pair
Shortlisted pairs are evaluated by the AI model, which returns a confidence score and a human-readable reason for each suggested match.
Review, Confirm, or Match Manually
Suggested matches appear with confidence badges and explanations. Confirm correct matches, reject false positives, or use the manual interface for anything the AI could not resolve.
Who Benefits Most
Designed for finance professionals and teams managing high-volume documents
AP and Accounting Teams
Reduce monthly reconciliation time by 8+ hours. The pipeline handles the matching; your team reviews only the flagged exceptions and edge cases.
Finance Controllers and Auditors
Every suggested match comes with an AI-generated reason and a confidence score — creating a documented, reviewable audit trail for each reconciliation decision.
Bookkeepers Managing Client Accounts
Reconcile dozens of bank statements against thousands of invoices across clients. AI scoring handles volume; confidence thresholds maintain accuracy.
See AI Transaction Matcher in Action
Set up in under 5 minutes and let AI handle the busywork.
Frequently Asked Questions
Stage 1 is an algorithmic prefilter that narrows candidates by currency (exact match), date window (-7 to +60 days from invoice date), and amount tolerance (within 10%). Stage 2 sends surviving candidate pairs to an AI model that scores each on vendor name, description, amount, and date — returning a 0-to-1 confidence score.
Only matches with an AI confidence score of 0.6 or higher are surfaced as suggestions. This threshold is calibrated to deliver a 92% auto-match rate while minimizing false positives.
A reconciliation record is created linking the transaction and invoice. The invoice status is automatically set to "paid." The match is permanent and visible in your reconciliation log.
Yes. Rejecting a match removes it from the suggestion list and excludes that pair from all future matching runs — so the same false positive will not reappear.
Unmatched transactions appear in a dedicated tab. From there you can use the manual matching interface to browse unmatched invoices and link them directly. Manual matches follow the same confirmation and auto-pay workflow.
No. Matching is included in the bank statement upload credit at no additional cost, regardless of how many transactions or invoices are compared.
Yes. Trigger a re-run from the dashboard at any time. New invoices will be matched against previously unmatched transactions, and new transactions against existing invoices.
Each suggestion includes a plain-English explanation — for example, "Vendor name 'Acme Corp' closely matches transaction description 'ACME CORP PAYMENT', amount difference is $0.50, dates are 3 days apart." This gives reviewers transparency into every match decision.
You Might Also Need
Complementary tools that extend this capability