Two Keys, One Payment: Rehearsing the Call That Stops a Diverted Invoice
Photo: Infoacrosseverything, CC BY 4.0, via Wikimedia Commons
Most UK small firms have trained their people to distrust links, attachments and unexpected requests. That training is worthwhile, and it will not save you here. The invoice fraud that succeeds in 2026 usually arrives inside a genuine conversation, from a genuine address, with the correct invoice numbering and the correct sign-off. Nothing looks wrong because nothing is wrong with the email except the account number at the bottom.
The control that stops it is not awareness. It is a procedure, jointly owned by finance and IT, that has been rehearsed out loud at least once.
The failure is procedural, not technical
A common pattern: a supplier's mailbox is compromised, the attacker reads the thread quietly for a fortnight, then replies to an outstanding invoice with revised bank details and an apology about a banking migration. The message passes authentication because it genuinely came from the supplier. Your mail filter has nothing to catch. Your staff see a familiar signature block and a plausible reason for the change.
At that point the only thing standing between the firm and a Faster Payment is a rule that says bank detail changes are verified out of band, by two people, before anything moves. Whether that rule exists on paper matters less than whether the people who would face the email know exactly what to do in the first five minutes.
Why finance and IT must sit in the same room
Finance owns payment approval and the supplier record. IT owns the mailbox estate, the sign-in logs and the mail flow history. Neither holds enough of the picture alone. A drill run only by the finance team tends to end with the right instinct and no evidence; a drill run only by IT tends to end with a list of log queries nobody has authority to act on.
Twenty minutes together, once a quarter, closes that gap cheaply.
Setting up: three roles, no live systems
You need a payer from finance, a second approver who does not normally release payments, and someone from IT who can read logs. One observer holds a timer and a notepad. Print a fabricated email, bring the supplier onboarding record for one real supplier, and have the payment run timetable to hand. Do not touch live systems during the exercise.
Minutes 0 to 5: the request lands
Read the email aloud. Ask three questions and write down the answers. Who would open it first? What would make them act quickly this week — month end, a director on leave, a supplier chasing? And what is the actual cut-off time for the next payment run?
You are looking for the moment where urgency beats verification. It is almost always present.
Minutes 5 to 12: the call that must be made
The rule is simple: verify using a telephone number held independently of the email. That means the number on a previous statement, in the contract, or in the supplier onboarding record. Not the signature block. Not the number quoted in the email. Not a mobile number offered in a reply.
Ask for the accounts contact by name and ask them to confirm the change in writing to a known address. Bank Confirmation of Payee is a useful prompt but not an approval: a fraudster's mule account can return a close or partial match, and a match means only that the name on the account resembles the one you typed.
Then apply separation of duties. The person who verifies cannot be the person who releases. A workable standing rule for small teams is a 48-hour hold on any change of bank details, a second call to a different contact, and director sign-off above a set threshold.
Minutes 12 to 17: the technical sweep
IT takes over. Even where the compromise sits with the supplier, your own estate can tell you how exposed the thread was. Check whether any finance mailbox has new inbox rules, forwarding or auto-replies that nobody recognises. Look for unfamiliar devices and unusual sign-in locations. Review delegated access to the finance mailbox. Search mail flow for the supplier's domain and for lookalike domains registered recently. Confirm the DMARC result on the received message.
One further check is worth institutionalising: has the same request gone to other customers of that supplier? Attackers routinely work a supplier's whole customer list. A warning to your own team and, if you can, to the supplier's other local contacts costs nothing.
Minutes 17 to 20: the debrief
Write down every point where the exercise stalled. In most firms it stalls at the same three places: no verified contact list, no agreed owner for bank detail changes, and nobody who knows the bank's fraud line.
The one-page script
Keep this beside the payment run and in the finance team's documentation:
- Stop. Do not reply to the email. Forward it internally as an attachment if needed.
- Call the supplier on the independently held number and speak to a named contact.
- Request written confirmation to a previously known address.
- If funds have already left, call your bank's fraud line immediately and quote the payment reference.
- Report the incident to Action Fraud and retain the reference.
- Notify your cyber insurer within the notification window in the policy.
- Preserve the email with full headers; do not delete the message or the thread.
- Brief every payment approver before the next run.
What the drill usually exposes
Three findings recur. First, the supplier record is a spreadsheet with a switchboard number and no named contact. Second, bank detail changes are handled over email because that is how the supplier prefers it. Third, when one person is on leave, the approver and the verifier become the same human being.
Each of those is fixable inside a week and none requires a budget line.
Keeping the rehearsal honest
Repeat the exercise when the finance team changes, when the payment platform changes, and after any genuine supplier bank detail amendment. Log the date, the findings and who closed them. A drill that nobody records becomes folklore within a year, and folklore is what fails when the invoice arrives.