Every pair has a reason
Rule matches show the rule. AI suggestions show a confidence score and a sentence such as "payment reference matches PO 4471 on invoice 10318, amount equal". Reviewers check the reason, not the arithmetic.
Reconciles is AI reconciliation software for US finance teams that match bank statements, processor payouts and invoice payments against a ledger. A deterministic matching engine pairs the clear cases first with a stated reason, a generative AI model then proposes pairs for the hard leftovers with a confidence score and a plain sentence, and nothing is accepted until a person approves it. Plans are published, from $888 a year, and you can see matches from your own files before you sign up.
Your files are processed in your browser and never stored. No card needed.
Most products sold as AI reconciliation tools are rules engines with a new label. SolveXia, a reconciliation vendor itself, says as much in its own analysis of AI reconciliation software: most of the market is still rule based, and the honest value of machine learning sits in the lines rules cannot pair. We agree, and Reconciles is built that way on purpose.
Exact amounts, invoice numbers in a memo, a payout that equals a bank deposit two days later: those are rules, and a rule is faster, cheaper and easier to audit than a model. The model is only asked about what is left, which is where finance teams actually lose hours. A customer pays three invoices with a reference that names none of them. A remittance shortens the payee to initials. A processor nets a refund and a fee into one deposit. That tail is where AI earns its place.
What the AI never does in Reconciles: post to your ledger, accept its own suggestions, or change amounts. Accepted pairs go into the report and into the journal entry file you import after review.
| Step | What does the work | What you see |
|---|---|---|
| 1. Exact and rule matches | The matching engine, on amount, date window, reference and payee | Paired lines with the reason, for example "invoice number in reference" |
| 2. Combinations and partials | The engine, searching open items that add up to the payment | One deposit shown against several invoices, or a partial with the balance left open |
| 3. Leftovers | A generative AI model reads the unpaired lines on both sides | Suggested pairs with a confidence score and a sentence explaining why |
| 4. Decision | You | Accept or reject each suggestion, one by one or in bulk |
| 5. Exceptions | The exceptions queue | Unmatched lines, short pays and duplicates with an owner and comments |
The engine handles one to one, one to many and many to one matching on its own, so the AI step stays small. That keeps the run fast and keeps the bulk of your matches explainable by a rule an auditor can read. The plain engine is described on the transaction matching software page.
Take a distributor with one operating account and an ERP export of open receivables. In October the bank file has 2,400 lines and the open invoice list has 2,150 items. This is an illustration of how a run splits, not a promise of your match rate.
| Stage | Lines handled | What happened |
|---|---|---|
| Rules, one to one | 1,880 | Same amount, invoice number or customer name in the memo, date within the window |
| Rules, one deposit to many invoices | 210 | ACH deposits that sum two to nine invoices of the same customer |
| AI suggestions accepted | 190 | References that name a purchase order instead of an invoice, payee names shortened by the bank |
| AI suggestions rejected | 35 | Plausible amounts from the wrong customer, sent back to the queue |
| Exceptions left | 85 | Short pays, an unapplied deposit and four duplicates, each with an owner |
On Growth, the 225 AI suggestions in that run use part of the 2,000 a month the plan includes. The controller reviews 225 proposals and 85 exceptions instead of 2,400 lines, and every accepted pair carries either a rule reason or the model's sentence. For the invoice side of this flow, see invoice reconciliation software.
| Reconciles | HighRadius | Simetrik | Trintech | Xero | |
|---|---|---|---|---|---|
| How you buy | Self-serve, by card | Sales led quote | Demo, then quote | Sales led quote | Subscription to the accounting system |
| Published price | $888 to $5,988 a year, or $149 to $999 monthly | No | No | No | Yes, as part of Xero plans |
| Where the AI sits | Suggestions for lines rules left unpaired, each approved by a person | Matching and cash application inside an order to cash suite | Agentic AI on a no-code platform for high volume payments data | Account reconciliation and substantiation inside a close suite | Bank feed match suggestions inside Xero |
| Data it reconciles | Bank, payout and invoice files against any ledger export | Enterprise receivables and account balances | Payments, fees, chargebacks and settlements for banks, PSPs and marketplaces | Balance sheet accounts for large corporates | Xero bank feeds against Xero transactions |
| Best for | US mid-market teams, ecommerce and SaaS finance, accounting firms | Large enterprises with SAP or Oracle | Payment companies with billions of records | Enterprise close teams | Small businesses already on Xero |
HighRadius advertises its own auto match and accuracy figures on its account reconciliation page, and every vendor in this table will tell you its AI is the most accurate. The only number that matters is your own match rate on your own month, which is why Reconciles shows it before you pay. Simetrik is compared in detail in our Simetrik pricing and alternatives guide, and the wider category in best bank reconciliation software.
Rule matches show the rule. AI suggestions show a confidence score and a sentence such as "payment reference matches PO 4471 on invoice 10318, amount equal". Reviewers check the reason, not the arithmetic.
Suggestions wait in a review list. You accept or reject them one by one or in bulk, and rejected lines go back to the exceptions queue.
Scale adds approvals before a run is final and an audit log of who accepted what, which is what an auditor asks for when AI touched a reconciliation.
Growth exports journal entries for QuickBooks and Xero, and Scale adds NetSuite, so fees and adjustments reach the ledger only after the pairs are approved. The bank reconciliation software page shows the full flow.
| Plan | Yearly price | Monthly price | Matched transactions a month | AI suggestions a month |
|---|---|---|---|---|
| Starter | $888 | $149 | 3,000 | 300 |
| Growth | $2,388 | $399 | 20,000 across 5 entities | 2,000 |
| Scale | $5,988 | $999 | 100,000 across 25 entities | 10,000 |
There is no implementation fee and no separate charge for the AI step. The rules engine has no allowance of its own beyond the matched transaction limit, so the AI allowance only counts the hard leftovers. Full details are on the pricing page.
Yes, with a person signing off. AI is good at pairing bank lines whose references are vague, shortened or point to a purchase order instead of an invoice. Clear matches are better left to rules, which are faster and easier to audit. In Reconciles the model suggests, explains and scores each pair, and a reviewer accepts or rejects it.
The best AI reconciliation tool is the one with the highest match rate on your own files at a price you can see. Enterprises with SAP often pick HighRadius or BlackLine, payment companies look at Simetrik, and mid-market teams matching bank, payout and invoice files use Reconciles from $888 a year.
It is as accurate as the review around it. Models can propose a plausible pair from the wrong customer, so a confidence score and a stated reason matter more than any headline accuracy figure. Reconciles never accepts its own suggestions, and rejected proposals return to the exceptions queue for a person.
No. AI removes the line by line matching, which is the slow part, and leaves the judgment: short pays, disputed fees, unapplied cash and what to book. The reviewer reads proposals and exceptions instead of every line, and spends the saved time on the items that need a decision.
Upload a bank statement and a ledger or invoice export. In a few minutes you will see what the rules match, what the AI proposes for the rest and what is left for you, before you pick a plan.
Your account is saved and you are logged in. Your result now shows one more page of matched rows. You also get one more sample run of up to 500 lines.
To reconcile full months, open your workspace with a plan.
Choose a planStarter plan selected
Your account is saved and you are logged in. Pay by card on the next page and your workspace opens right away.
$74 per month, billed $888 yearly
$149 per month, billed monthly
Growth plan selected
Your account is saved and you are logged in. Pay by card on the next page and your workspace opens right away.
$199 per month, billed $2,388 yearly
$399 per month, billed monthly
Scale plan selected
Your account is saved and you are logged in. Pay by card on the next page and your workspace opens right away.
$499 per month, billed $5,988 yearly
$999 per month, billed monthly
Your account is saved and you are logged in. You can now run one sample reconciliation of up to 500 lines.
To reconcile full months, open your workspace with a plan.
Reconcile now Choose a planEvery plan includes the matching engine, the exceptions queue and exports. Pay yearly and save 50%.
Starter
$74 per month, billed $888 yearly
$149 per month, billed monthly
Growth
$199 per month, billed $2,388 yearly
$399 per month, billed monthly
Scale
$499 per month, billed $5,988 yearly
$999 per month, billed monthly
Cancel any time. Your plan stays active until the end of the paid period.