When the invoice is real and the bank details are not

Finance manager phoning a supplier to verify changed bank details before releasing an invoice payment

A supplier you have paid for years sends an invoice. The project reference is right, the amount matches work that was actually done, it comes from the address you always use, and it lands in a thread the two of you have had running since spring. At the bottom there is a new account number and a line saying they have moved banks.

Nothing in that message has been forged. Somebody else has the supplier’s mailbox open and is writing from inside it, which is why the checks most firms rely on do not raise anything.

Why this is harder than ordinary invoice fraud

Ordinary invoice fraud arrives from a lookalike domain, so there is something to notice: a missing letter, a domain registered the week before, a display name that disagrees with the address underneath. Mail filtering can score a message like that, and a careful reader can catch it.

The FBI’s Internet Crime Complaint Center describes business email compromise as a scam targeting businesses working with suppliers, and businesses regularly performing wire transfer payments, carried out by compromising email accounts through social engineering or computer intrusion techniques. Because the account itself has been taken, by the time the invoice reaches you there is no forgery left to detect.

In its 2025 Internet Crime Report the Internet Crime Complaint Center recorded 24,768 business email compromise complaints and just over three billion dollars in reported losses, within a total of 20.9 billion dollars across every crime type it tracks.

What happens before the invoice arrives

Somebody who gets into a mailbox usually sends nothing for a while. They read the billing history, work out who approves payments and what the invoice cycle looks like, and in some cases add a rule that moves replies into a folder the owner never opens. When the request finally goes out it is timed to a real invoice and carries figures that match.

Your own defenses are pointed somewhere else throughout. Domain authentication passes, because the domain is genuine and the message was sent from it, and multi-factor authentication protects a tenant that nobody is attacking. Training that teaches people to spot a fake sender does not help a great deal when the sender is real. The manipulation techniques that get somebody into a mailbox in the first place are covered in our write-up on social engineering beyond phishing.

How does this reach an architecture, engineering or construction firm?

Payment on a project is irregular by nature. Progress billing, retention released late, change orders that move the numbers halfway through, a consultant who invoices once a quarter. A new account number on a genuine invoice has very little to stand out against.

Public schedules, subcontractor lists and award notices also make it straightforward to time a request to work that is under way. Our construction practice page sets out what that pattern looks like on a job site.

The control that works

Telephone the supplier and ask whether they changed their bank details, using a number you held before the request arrived.

The FBI’s guidance on business email compromise says to verify any change in account number or payment procedures with the person making the request, and to look up the company’s phone number yourself instead of using the one provided. Where the number comes from matters as much as the call does. A number in the signature block, in the invoice PDF or in the body of the message can be edited by whoever is sitting in the mailbox, so take it from your accounting system’s supplier record or from the signed contract. If the only number you hold arrived with the request, there is no verification step to perform.

What to decide before it happens

Firms that handle this well settled the rule in advance, at a point when nobody was in a hurry. Four things belong in writing:

  • Who can authorize a change to supplier bank details. Name the roles. A change approved by whoever happened to be at their desk has not been approved in any meaningful sense.
  • What verification means. A call to the number on file, to a named person, logged with the date and who made it.
  • A threshold that brings in a second pair of eyes. Pick a figure that matches how you actually pay, and apply it to new account numbers as well as to new suppliers.
  • A standing instruction that nobody is penalized for the delay. Put it in writing alongside the other three, so that holding a payment for two hours is a decision the procedure already backs.

Tell your own suppliers that the rule runs in both directions, so a callback from you is expected and one made to you gets answered.

What if the supplier confirms the change by email?

The confirmation arrives from the same mailbox as the request, so it adds nothing. Whoever is reading that account can answer a question about the invoice as convincingly as they sent it. A confirmation is worth something only when it comes back on a channel the request did not use, which in practice means a telephone call you placed yourself. An attached letter on the supplier’s headed paper carries the same problem, since the attachment comes from the same place.

Who should make the callback?

Somebody other than the person who received the invoice, if you have the people for it. Anyone who has already read a request and found it reasonable will tend to read it the same way the second time. In a small finance team one person doing both is still worth doing, provided the number comes from your own records and the call gets logged with a date and a name.

The first hour, if a payment has already gone out

Call your bank and ask for a recall on the transfer, describing it as fraud rather than an error. Report it to the Internet Crime Complaint Center, which runs a process for freezing fraudulent domestic wire transfers and works better the sooner it hears. Telephone the supplier as well, because their mailbox may still be under someone else’s control and an emailed warning may never reach them. Keep the original message and its headers before anybody starts tidying up.

Afterwards it is worth checking whether your own mailboxes were involved, since a compromise sometimes runs in both directions through a shared thread. Our cybersecurity services cover that review, along with the mailbox rule audit that finds the forwarding and deletion rules an intruder leaves behind.

Where this sits next to your cyber insurance

Funds transfer fraud is usually assessed separately from ransomware on a cyber liability application, and carriers ask about the verification step directly. A written callback procedure with a log showing that it runs is a stronger answer at renewal than a policy nobody can evidence.

If you are not certain what your current process would do with a supplier’s new bank details on a real invoice, that is worth half an hour with somebody who sees this across a lot of firms.

Want this looked at properly?
A first conversation takes fifteen minutes. We will tell you what we would look at first, and what it would cost.