Nine Invoice Fraud Prevention Mistakes Finance Teams Make, and What to Do Instead
Most diverted-payment losses trace back to a handful of habits that recur across accounts-payable teams. Here are nine, the control that replaces each one, and why your insurer may already expect it.
Callback Mistakes That Let Fraudsters Verify Themselves
Say a long-standing supplier emails your accounts-payable team today to announce new bank details. Could the team prove the request is genuine before the next payment run? Most finance teams would say "we'd call them." That's the right instinct. The problem is who they'd call, and it's the point where most invoice fraud prevention programs fail.
Mistake 1: Calling the number in the email, invoice or letterhead. Once a fraudster controls the message, they control every contact detail in it. A new phone number in a signature block or a "new remittance contact" on a PDF invoice routes your clerk straight to the attacker, who will gladly confirm the new account.
What to do instead: Call back only on a number already in the vendor master file, sourced from the signed contract, the original onboarding form or the supplier's public main switchboard. Never use a number supplied in the request itself. Write that rule into the AP procedure verbatim. Insurers' verification conditions often refer to a pre-established or previously known contact, and matching that language makes it easier to show you met it.
Mistake 2: Treating a reply from the vendor's real email domain as proof. Many cases now start inside the supplier's own compromised mailbox, usually called vendor email compromise. The sender address is genuine. So is the thread history. The attacker may have been reading along for weeks, so replying "please confirm" in that thread gets you an answer from the attacker.
What to do instead: Treat every bank detail change as unverified until it's confirmed through a separate channel. Train staff to spot lookalike domains (a swapped letter, an extra hyphen, a different top-level domain), then teach the harder lesson: a perfect domain proves nothing.
Change Control Failures Where One Approver Becomes the Weak Point
The next three mistakes are about process design, not staff vigilance. Each one puts too much trust in a single person or a single moment.
Mistake 3: Letting one person receive, verify and approve a bank detail change. If one AP clerk can open the request, make the callback and update the vendor record, one bad call loses the money. It also leaves a dishonest insider completely unchecked.
What to do instead: Split the change into three roles: one person logs the request, a second makes the independent callback, and a third (usually an AP manager or controller) approves the change in the system. Enforce this in your ERP or AP platform, not just in a policy document. Mainstream systems such as SAP, Oracle NetSuite, Microsoft Dynamics 365, Sage Intacct and Xero support approval workflows or audit trails on vendor master changes, though capabilities vary by edition and configuration. A 15-person finance team can usually run three-way separation. Smaller teams should borrow an approver from outside finance, such as the CEO or an operations lead.
Mistake 4: Paying the next invoice in full right after a change. Fraudsters time their requests to land just before large payments fall due.
What to do instead: Hold the first payment to any changed account for a few business days (three to five is a sensible default), and use payee-name checks wherever your payment rails offer them. The UK's Confirmation of Payee, the EU's Verification of Payee for euro credit transfers, beneficiary name lookup for NEFT and RTGS in India, and the recipient-name display on Brazil's Pix can all surface a mismatch. Coverage and reliability differ by country and bank, and many cross-border SWIFT payments have no equivalent check. Treat these tools as an added layer, never a substitute for the callback.
Mistake 5: Letting urgency or seniority override the process. "The CFO needs this out before the close" is the oldest script in business email compromise.
What to do instead: Have the CFO and CEO formally adopt a policy stating that no one can waive verification, themselves included. Staff hold the line under pressure only when the most senior people have publicly agreed to be bound by it.
Deepfake Requests and the Limits of Email Filters
Mistake 6: Acting on a voice or video request because it sounds and looks right. Voice cloning and real-time video impersonation are now cheap and convincing enough that a familiar voice no longer counts as verification. A two-minute call from what sounds like your group CFO, asking for a confidential acquisition payment, is a well-established attack pattern.
What to do instead: Treat voice and video like email: channels that can be spoofed. For any payment instruction arriving by phone, video meeting or messaging app, hang up and call back on the internal directory number. Above a set threshold, require a second approver in the payment system. Mid-size companies often set that threshold somewhere in the low five figures (USD), but size it to your typical payment values. Some teams add a rotating code phrase for executive payment requests. It costs nothing, but it fails the moment the phrase appears in email or chat.
Mistake 7: Assuming your email security tool is your business email compromise protection. Filters matter. Enforce DMARC at p=reject on your own domains, require phishing-resistant MFA on finance mailboxes, and alert on new mailbox forwarding rules. None of that stops a message sent from a supplier's real, compromised account.
What to do instead: Run technical and process controls side by side, then decide whether a dedicated payment-verification layer earns its cost. The figures below are typical 2026 market ranges; they vary by contract, user count and region, so confirm with vendors.
| Control layer | Examples | Indicative cost (USD, 2026) | Stops a compromised-vendor email? |
|---|---|---|---|
| Advanced email security | Microsoft Defender for Office 365, Proofpoint, Mimecast, Abnormal Security | Roughly $2-$8 per user per month, depending on tier and bundle | Sometimes, through behavioral anomaly detection |
| Payment and vendor verification platforms | Trustpair, Eftsure and similar tools | Usually quote-based; mid-size deployments commonly run into five figures a year | Often, by checking account ownership before payment |
| Process controls (callback, dual approval, hold period) | Internal procedure plus ERP workflow | Mainly staff time | Yes, when followed every time |
For companies paying no more than a few hundred suppliers, well-enforced process controls inside the ERP give the best return. A verification platform gets easier to justify with high supplier turnover, many cross-border payees, or AP teams operating in several countries.
Crime Insurance vs Cyber Insurance: Where Diverted Payments Fall Through
Mistake 8: Assuming your cyber policy pays when an employee is tricked into sending money. Many traditional crime and fidelity policies exclude losses where an employee knowingly authorized the transfer, commonly through a voluntary parting exclusion. Some cyber policies exclude first-party fund losses altogether. Social engineering cover is typically added by endorsement, with a sublimit well below the main policy limit.
| Loss scenario | Commercial crime policy (typical) | Cyber policy (typical) |
|---|---|---|
| Employee pays a fake invoice after a spoofed email | Usually only with a social engineering fraud endorsement | Sometimes, through a social engineering or cybercrime sublimit |
| Attacker breaks into your bank portal and sends funds | Often covered under computer or funds transfer fraud | Varies; often covered |
| Your customers pay a fraudster posing as you | Rarely covered | Sometimes covered as "invoice manipulation" |
| Forensics and notification after a mailbox compromise | Generally not covered | Generally covered |
For small and mid-size buyers, social engineering sublimits are often in the low six figures (USD). Higher limits exist, but underwriters scrutinize your controls harder before offering them. Wording, availability and regulatory treatment differ significantly across markets such as the US, UK, EU, Singapore, the UAE and India, so have a licensed broker map your actual policies against these scenarios.
Mistake 9: Following the control but keeping no record of it. Social engineering endorsements frequently make cover conditional on verifying payment changes by callback or another independent method. Some proposal forms ask about this directly, and an inaccurate answer can jeopardize a claim. If you can't show the callback happened, proving you met the condition gets much harder.
What to do instead: Log every verification: the date, who called, the number used and where it came from, and who approved the change. Keep the log in the ERP or a locked register, not someone's inbox. That record is what turns a good control into one you can claim against.
A 30-Day Plan to Make These Fixes Insurable
None of this needs a major project. A finance team can stand up the core of an invoice fraud prevention program in a month:
- Week 1: clean the vendor master. Confirm every active supplier has a contact number from an independent source, and flag any bank details changed in the past 12 months for review.
- Week 2: write the procedure. Cover callback sourcing, three-role separation, the first-payment hold, the executive no-override rule and voice/video handling. Have the CFO sign it.
- Week 3: enforce it in the system. Configure ERP approval workflows, set dual approval in banking above your threshold, and switch on payee-name checks wherever your banks offer them.
- Week 4: test it and send it to your broker. Run a tabletop exercise with a fake bank-change request and a simulated executive voice call. Then give your broker the written procedure and a sample verification log ahead of renewal.
Apply one rule to every payment: if AP can't show, from a record made at the time, that a bank detail change was confirmed on a number your company already held and approved by someone other than the person who received it, the payment doesn't go out. That rule protects the cash, and it's the one your insurer is most likely to test when you claim.
Quick answers
Is a small test payment enough to verify new vendor bank details?
No. A test payment proves only that the account exists and accepts money, not that your supplier owns it. The fraudster controls that account and will happily confirm the test amount arrived. If you use test payments at all, add them on top of an independent callback and a payee-name check, never in place of them.
Should we buy social engineering cover under our crime policy or our cyber policy?
That depends on how your existing policies are worded and where the gaps sit. Crime policies often give clearer funds-transfer and employee-theft cover, while cyber policies typically handle incident response costs after a mailbox compromise. The priority is avoiding gaps and overlapping sublimits, so ask a licensed broker to test both policies against specific diverted-payment scenarios before deciding.
Will an insurer deny a claim if we skipped the callback once?
It can, if the policy makes cover conditional on verifying payment changes and your records can't show it happened. How strictly that applies depends on the policy wording and local insurance law, which vary by country. Have a broker or coverage counsel review your specific conditions, and keep verification logs so the question never arises.
Written by the Auditinsure Direct editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.