Match what landed in the bank to what you invoiced, and get a clear explanation for every difference. Exact matches settle themselves, part payments stay visible, and the odd amount that matches nothing gets flagged with a best guess instead of sitting in limbo for a month.
The Invoice Tracker is a live Taskade Genesis app you can run on this page. It holds the invoice side of a reconciliation already, so that is the half you do not have to build; the payment feed and the matching rules are what you add. Try it, then click "Use this app" to clone and own a copy in about ten seconds and build the matching described below on top of it.
The build gives you a matching engine with judgement. The agent works through amount, reference, client name, and date proximity, then classifies each payment as an exact match, a probable match needing one click, a part payment, an overpayment, or unidentified. Differences are explained rather than reported: a payment 2.9 percent light on an international transfer is almost certainly a bank fee, one exactly matching two invoices is a combined payment, and one from a name you do not recognize is often a client's parent company. Every decision stays visible so the record shows how a match was made.
Workspace DNA is why matching gets better over time instead of staying equally tedious. Memory holds how previous payments from each client behaved, including the ones that always arrive short by the same bank fee. Intelligence, powered by 15+ frontier models from OpenAI, Anthropic, Google, and open-weight providers, uses that history to propose a match with a reason attached. Execution settles the clear cases and updates every downstream record instantly.
The AI agents inside the app carry 34 built-in tools, and the ones that matter here are persistent memory, file analysis for reading a bank export, and structured reasoning that shows its working. Every match records why it was made, which means someone reviewing it in six months can follow the logic rather than trusting it blindly.
The real gain is that unpaid invoices stop hiding behind bad data. Most businesses that think they have a collections problem partly have a reconciliation problem: invoices marked paid that were not, and payments received that were never matched. Fixing the matching often shrinks the chase list on the first run.
Work it in whatever view is fastest:
- Table view for the matching grid with payments against invoices
- List view for the queue of items needing a human decision
- Board view to sort into matched, part paid, and unidentified
- Calendar view to see payments landing over time
- Plus the rest of the 7 project views
Automations remove the routine half. A Taskade automation settles exact matches automatically, holds the ambiguous ones for you, and updates invoice status and client balances the moment a match is confirmed. Across 100+ bidirectional integrations, Stripe and Google Sheets pull payment data in while Slack and Gmail push confirmations and queries out. See filter data for the matching rules.
Clone it, invite whoever handles the bank, and the reconciliation queue stops being someone's Friday afternoon.
A necessary boundary: this reconciles payments to invoices as an operational task. It is not bank reconciliation in the accounting sense and it does not replace the reconciliation your accountant performs against statements. It makes the numbers you hand over already agree.
Run the first pass over three months of history rather than starting from today. It takes a few minutes, it usually surfaces two or three payments that were never matched to anything, and it gives the agent enough of each client's behaviour to make its future suggestions genuinely useful rather than generic.
One thing to decide up front is how much automation you want. Settling exact matches automatically is safe and saves most of the time. Settling probable matches automatically saves a little more and occasionally creates a mess, so most teams leave that step with a human and never regret it.
See finance workflows in the Community Gallery, start yours at Taskade Genesis, or read finance apps. It keeps Build an Invoice Status Tracker truthful, and it settles the installments created in Offer Installment Payment Plans.
Frequently Asked Questions
What if a client pays several invoices in one transfer?
The agent looks for combinations that sum to the payment amount and proposes the split, showing which invoices it believes are covered. You confirm, and all of them settle together.
How does it handle payments with no reference?
It matches on amount, client name, and timing, then presents its best guess with a confidence note. Unidentified payments go into a queue rather than being forced into a wrong match, which is far cheaper to fix.
Can it handle currency differences?
Yes. Where an invoice was issued in another currency it compares against the expected converted amount and attributes the gap to rate movement or fees. See Invoice International Clients in Multiple Currencies.
Does it work with bank transfers as well as card payments?
Yes. Card payments arrive as events and can settle automatically. Bank transfers usually arrive as a statement export, which you paste or import, and the same matching logic applies.
What happens to an overpayment?
It is flagged with the excess amount and held as a credit against the client, so it can offset the next invoice or be refunded. See Handle Refunds and Credit Notes.
How often should I reconcile?
Weekly is enough for most service businesses, and daily if cash is tight. A scheduled automation can run the matching and leave you only the exceptions.
Will it catch an invoice marked paid that never was?
Yes. Invoices with no matching payment are exactly what the exception report surfaces, and that error is more common than most teams expect.
