Bank Agent
The most expensive fraud in accounts payable is not a fake invoice. It is a real invoice, from a real vendor, paid into an account that changed last week.
A beneficiary account cannot change silently between approval and payment.Bank details are verified against what was approved, not against what is current in the vendor master — which is the exact gap a vendor-impersonation attempt is built to exploit.
What it checks
On every invoice, not a sample.
- Beneficiary account against the one on the approved entry
- Any change to vendor bank details since approval, and who made it
- Account name against the vendor name
- First payments to a newly changed account, held for confirmation
What it reads, what it writes
Reads
The approved entry and its bank details · Vendor master change history · The payment file before release
Writes
A hold where details changed after approval · The change, with who made it and when · Confirmation before the first payment on new details
Where it sits in the team
Agents do not work alone. Each one hands its result to the next, so a finding raised here shows up as context downstream rather than being re-derived.
Hands to it
It triggers
When it isn't sure
A legitimate bank change looks identical to a fraudulent one at the moment it is made. So the agent never decides which it is — it stops the first payment on changed details and asks for confirmation through a channel that is not the one the change arrived on.
Any check that runs on every transaction will meet cases it cannot settle. What makes a control trustworthy is not that it never hesitates — it is that hesitation produces a named question for a person, rather than a silent pass or a silent block.
See what this one finds in your last 90 days.
One export. No connector. We come back within a working day.
$Check your savings→