
Positive Pay Says No. Now What?
The alert lands mid morning with a deadline of early afternoon. Pay or return. Most teams decide on instinct because the evidence lives in four systems. It does not have to.
Positive pay is a bank service that matches every cheque or ACH debit presented against a file of items the company issued, and holds anything that does not match. The exception is the item the bank could not match, and the company usually has only a few hours to decide pay or return. The control fails not at the bank but at the review, where the person deciding often cannot see the issuing record, the invoice and the vendor history in one place.
- Positive pay works. The weak point is the exception review, which runs on a same day clock with evidence spread across the bank portal, the ERP, the AP file and the vendor master.
- Most exceptions are benign: a manual cheque outside the issue file, a payee name formatted differently, a reissued item. A small number are not.
- Deciding by default is the real risk. Teams that pay everything under a threshold have effectively turned the control off for the amounts fraudsters most often use.
- The evidence needed is always the same four items, which means the assembly can be automated even though the decision cannot.
- ChatFin pulls the issue record, the underlying invoice, the vendor bank history and the prior payment pattern into one view before the deadline, and logs the decision rationale.
Positive pay is one of the few bank controls that genuinely stops fraud before the money leaves. It is also one of the least well operated.
The service itself is simple. The company sends the bank a file of every cheque or authorised debit it issued. The bank matches what gets presented against that file and holds anything that does not match, usually reporting it as an exception the same morning it is presented. The company then has a short window, often ending early afternoon, to say pay or return. If nobody answers, a default applies.
That default is where the control quietly stops working. This page covers what an exception actually means, what evidence a good decision needs, why the deadline defeats most teams, and how to make the review something a person can complete properly in minutes.

What the bank is actually telling you
A positive pay exception is a mismatch, not an accusation. The bank compared a presented item against your issue file and something did not line up. The category of mismatch tells you a great deal about how worried to be, and it is usually reported but rarely read.
Amount and payee mismatches deserve the most attention because they are the alteration patterns. Item not on file is the most common and the most often benign, which is precisely why it is the one abused.
Why the review fails on a deadline
The exception arrives with a serial number, an amount, a payee string and a mismatch reason. Deciding well needs considerably more than that, and every additional piece sits somewhere else.
The issue record
Was this item issued at all, by whom, on what date, against which payment run? Manual cheques are the ones missing from the file, and finding out who wrote one takes phone calls.
The underlying obligation
Is there an approved invoice or expense behind this amount, and does the amount tie exactly? An exception with no supporting invoice is a different conversation entirely.
The vendor record
Has this vendor been paid before, at this amount, to this account? A first time payee on an exception item is the highest risk combination there is.
The recent change history
Was the vendor bank detail or address changed in the last ninety days? A recent change plus an exception is the pattern behind most successful payment fraud.
None of that is hard to gather. It is just spread across the bank portal, the ERP payments module, the AP document store and the vendor master, and the clock is running. So the reviewer does what anyone would do under time pressure and decides on the amount alone.

Deadlines force instinct
A pay or return decision due by early afternoon gets made on instinct unless the evidence is already at hand.
I check my Slack to make sure nothing's blowing up, X, Y, Z.
Caitlin Haberberger, on the AgentCFO podcast
Keep the issue file and vendor history one click from the alert, and set a standing rule for items you cannot verify in time.
The threshold habit that turns the control off
Almost every team develops an informal rule: pay anything under a certain amount, investigate above it. It is understandable and it is exactly the behaviour a competent fraudster tests for.
Amounts that clear without review are the amounts worth attempting repeatedly. The pattern in cheque and ACH fraud is not one large item, it is a sequence of unremarkable ones that each sit under whatever threshold the target uses, spread over enough weeks that no single month looks unusual.
The fix is not a lower threshold, which just moves the problem and buries the team. The fix is making the evidence cheap enough to gather that reviewing every exception properly becomes possible inside the deadline.
Where the tooling sits today
Bank portals handle the mechanics. Every major cash management bank offers positive pay with a same day exception workflow, and the file transmission side is well solved through standard formats and treasury connectivity.
Payment platforms and treasury systems sit above that. Bottomline and Modern Treasury manage payment origination and file delivery, Kyriba and GTreasury hold the bank connectivity, and vendor verification tools such as Trustpair address the adjacent risk of the account details being wrong before a payment is ever issued.
Nothing in that stack assembles the four pieces of evidence into one view on a deadline. That assembly is the actual work, and it is what ChatFin does across the ERP and the document store, in the same way it handles exceptions in NetSuite AP automation agents.

Making the decision defensible afterwards
Auditors and insurers ask a narrow question about positive pay: show me the exceptions for the period, the decision taken on each, who took it, and the basis. Most teams can produce the first two and struggle with the last.
A decision log that records the evidence considered turns the control from a claim into something demonstrable. It also changes behaviour, because a reviewer who knows the rationale is recorded is markedly less likely to pay on instinct.
| Exception record | Date, serial, amount, payee, mismatch reason as reported by the bank |
| Evidence gathered | Issue record found or not, invoice reference, vendor payment history, recent bank detail changes |
| Decision | Pay or return, the person who decided, the timestamp against the bank deadline |
| Rationale | One line on why, which is the field that separates a control from a habit |
| Follow up | Where returned, whether a fraud report was raised and whether the vendor was contacted |
Keep the log for the same period you keep bank reconciliations. It is the evidence that the control operated, not just that it existed.
How ChatFin assembles the exception view
ChatFin connects to the ERP you already run and to the document store where invoices and payment support live. When an exception is raised, the agent assembles the four evidence items against that serial and amount before the reviewer opens it.
The document processing agent locates the supporting invoice and reads the amount, vendor and approval from it. The autonomous reconciliation agent matches the presented item against issued payments and open items. The pattern recognition agent checks the vendor payment history and surfaces any bank detail or address change in the preceding window, which is the single most useful signal in the whole review.
| Document processing | Finds and reads the invoice or expense behind the presented amount |
| Reconciliation agent | Matches the item against issued payments, open payables and prior clearing |
| Data query agent | Answers follow up questions on vendor history in plain English during the review |
| Pattern recognition | Flags first time payees, recent bank detail changes and repeat amounts below threshold |
| Governed write back | Records the decision and rationale, and prepares any resulting ledger entry for approval |
The decision itself stays where it belongs. ChatFin does not pay or return anything. It puts the evidence in front of a person inside the bank deadline and records what they concluded, which is the same human in the loop pattern used across pre built finance AI agents.

Tightening this in one close cycle
This is one of the few control improvements that can be done inside a month without a project.
Pull last quarter's exceptions
Export every exception the bank raised, with the mismatch reason. Most teams have never looked at these in aggregate.
Categorise them
Split by mismatch type and by whether the decision was pay or return. The share of item not on file tells you how much manual cheque activity sits outside your issue file.
Fix the issue file first
Every manual cheque outside the file is a self inflicted exception. Getting those into the file removes most of the volume before you improve anything else.
Add the evidence view and the log
Assemble the four evidence items automatically and record the rationale. Only then consider whether the threshold habit is still needed.
Teams that run this sequence usually find exception volume falls by more than half from step three alone, which makes reviewing the remainder properly a realistic ask.
Amounts that clear without review are exactly the amounts worth attempting repeatedly. The threshold habit is not a shortcut, it is a published weakness.
ChatFin
Standard, Payee and Reverse Positive Pay Explained
Standard positive pay compares the check number and amount against your issue file. Payee positive pay adds the payee name to the comparison. Reverse positive pay flips the direction: the bank sends you the items presented that day and you check them against your own records. ACH debit filters do a similar job for electronic debits, using a list of originators you have approved.
Whatever the variant, an unmatched item becomes an exception with a cut-off time. Each bank sets its own default for a missed deadline, either pay or return, so read yours in the service agreement before you need it. Most exceptions are ordinary, like a manual check or a reissue. The ones that need care are items with no invoice behind them.
Chris Ortega, fractional CFO and CEO of Fresh FP&A, on who is responsible when AI is involved.
“There's never gonna be a case where the AI is the defendant. ... It is gonna be a person. It is gonna be somebody responsible for something.”
Positive Pay Exceptions You Will See in Practice
Each case below shows up in the bank portal as a mismatch, and each needs a different check before the pay or return click.
- A manual check missing from the issue file. Someone cut a check by hand for a deposit refund and never sent it to the bank. The reviewer asks who wrote it and finds the approval before paying.
- A payee name that reads differently. The bank read the name on the item as a slightly different spelling of the vendor. The reviewer compares it with the vendor master and the invoice, and returns the item if the names point to different parties.
- A reissued check after a stop payment. The original was lost and replaced, but the new serial was not sent in the file. The reviewer matches it to the stop request and updates the issue record.
- An ACH debit from an originator nobody recognizes. Treasury checks whether a cancelled subscription or an old authorization explains it. If not, the debit is returned and the originator is added to the block list.
- A repeat exception from the same process. At month end the team lists which exceptions recur and why. Moving a frequent manual-check payee to ACH removes the cause.
Caitlin Haberberger, CFO of Verdant, on risk and cash.
“There's just the overall risk aversion layer on top of all of this. If cash goes missing, if something goes undetected ... you're the check at the company on some of that risk.”
Quick Answers on Positive Pay Exceptions
What is positive pay and how does it work?
Positive pay is a bank fraud control. You send the bank a file of the checks or debits you issued, and the bank holds any presented item that does not match. You then tell the bank to pay or return each held item.
What happens if you miss the positive pay deadline?
The bank applies the default in your agreement, which is either to pay or to return the item. Know which one yours is. A pay default means a fraudulent item can clear if nobody answers in time.
How does reverse positive pay work?
The bank sends you a list of the items presented against your account each day. You compare it with your own issue records and tell the bank which to pay. The matching happens on your side instead of the bank's.
What should you check before returning a positive pay exception?
Find out whether the item was issued, whether an approved invoice supports the amount, and whether the vendor and its bank details have a history. If you cannot confirm any of these before the cut-off, record why you returned it.
How do you reduce positive pay exceptions?
Send the issue file after every payment run, include manual checks and reissues the same day, and fix payee names at the source in the vendor master. Fewer surprises mean fewer exceptions to review under time pressure.
Does positive pay stop check fraud?
It stops altered and counterfeit items from clearing unnoticed, as long as someone answers exceptions properly. It does not help if staff approve items by habit or the issue file itself is incomplete.
Frequently Asked Questions
What is a positive pay exception?
It is an item presented to the bank that did not match the issue file the company sent. The mismatch may be that the item is not on the file at all, that the amount differs from the issued amount, that the payee name does not match under payee positive pay, or that the serial number is duplicated or out of range. The bank holds the item and asks the company to decide pay or return, usually within the same business day.
How long do you have to respond to a positive pay exception?
Most banks report exceptions in the morning and require a decision by early afternoon on the same business day, though the exact cutoff varies by bank and service. If no decision is given, a default applies, and that default is set in the service agreement. Knowing which way your default runs matters, because a default to pay means silence is an approval.
Can AI decide positive pay exceptions automatically?
It should not, and ChatFin does not. The decision carries fraud and liability consequences that belong with a person. What can be automated is the evidence assembly: finding the issue record, the supporting invoice, the vendor payment history and any recent bank detail change, and putting them in one view inside the bank deadline. ChatFin then records the decision and rationale for the audit trail.
Does positive pay cover ACH as well as cheques?
Cheque positive pay is the original service and payee positive pay extends it to the payee name on the item. ACH positive pay, sometimes sold as ACH debit filtering or blocking, applies the same idea to electronic debits by allowing only authorised originators and amounts. Many organisations run cheque positive pay and leave ACH filtering unconfigured, which leaves the faster channel less protected than the slower one.
See ChatFin assemble a positive pay exception before the bank deadline.
ChatFin pulls the issue record, the supporting invoice, the vendor payment history and any recent bank detail change into one view, so the reviewer decides on evidence rather than on instinct.
The decision stays with your team. ChatFin records what was decided, by whom and on what basis, giving you the exception log auditors and insurers ask for.
Book a DemoBring last quarter's exception file.
Send an anonymised export of the exceptions your bank raised and ChatFin will show the evidence view it would have assembled for each one, inside the same day window.