Solutions · Banking · Fraud and money laundering
A business customer pays AED 480,000 to a supplier account from a fake e-mail, and within the hour the money moves through three mule accounts. Fraud, compliance and the login log each see part of it. On a knowledge foundation it is one case within minutes, an AI agent does the legwork inside the limits the bank sets, and the people with the authority decide.
Banking · Fraud and money laundering
Illustrative demonstration with a fictional bank, fictional customers, accounts, people and policies.
Nineteen forty. The phone rings at the fraud desk. A business customer paid four hundred and eighty thousand dirhams an hour ago, to a supplier account from a fake e-mail. Mariam, the analyst on duty, takes the call. Mariam can see the payment. Compliance has a separate alert: money in and straight out, with two working days to review it. A login log shows one device on three accounts. Nobody has the whole picture. On Samture's foundation, Mariam sees one case within minutes. The money moved through three mule accounts on one device, and two hundred and fifty thousand dirhams are still in the bank. An AI agent does the legwork, within limits the bank has set. Mariam approves each step, and the agent may only read. It wanted to call the customer on the number in the fake e-mail. Blocked: only the number on file. Every finding cites its policy clause and version. One without is sent back. The agent also spotted an earlier payment from another customer, who has said nothing. Then the people decide. Mariam sends the recall. Omar, the MLRO, restricts the accounts and decides on the report to the FIU. No one tips off the account holders. Eighteen minutes from the call to the restriction. The agent was fast. Mariam and Omar stayed in charge.
Why fraud and money laundering stay apart
Fraud operations works the victim's report. Transaction monitoring raises its own alert on the account that moved the money, with its own review deadline. The digital banking log holds the device that links the accounts, and nobody reads it as a case. By the time the alerts meet, the money has left.
Each system is right about its own part. None of them shows the mule network, so the case is put together by hand, if at all.
Money moves on within the hour. A restriction that waits for the next alert review comes after the funds have gone.
An assistant that can read every account and call any number can be steered by the fake e-mail itself.
What the foundation does
The foundation links the payment, the accounts, the device and the alerts into one case, without replacing the fraud, monitoring or core banking systems. An AI agent works the case with a fixed list of tools: a person approves each step, it may only read, and it stays inside the case. A call-back goes only to the number the bank has on file. Every finding cites the clause and version of the policy it rests on, and a finding without one is sent back. Nothing with an effect outside the case is written until the person with the authority signs: the fraud analyst for a recall, the MLRO for restricting accounts and for the report to the FIU.
The way in is the same as for every Samture engagement: a structural assessment first, then a demonstrator on your own processes and data, then a roadmap. Every engagement has one accountable Samture lead, from the first assessment to production.
Questions
No. It reads, links and drafts. Recalls, restrictions, contact with other customers and suspicious transaction reports are signed by the people who hold that authority in the bank, and each signature is recorded with its time.
Contact with anyone in the network is a separate decision of the MLRO, and the holders of the accounts under suspicion are not contacted. That rule is part of the policy the agent works against, so a draft that breaks it is held.
Content from the case, including the e-mail, is treated as data, never as an instruction. The agent can only use the tools on its list, every use passes a control point, and a call-back to a number that is not on file is blocked.
No. Those systems keep raising their alerts. The foundation connects what they see into one case and holds the bank's rules, so analysts and agents work from the same picture.
With the bank, on infrastructure it controls, in the UAE. The demonstration uses fictional data only.
With one recurring case type, such as payments to changed bank details. A structural assessment maps the systems and rules involved, followed by a demonstrator on the bank's own policies and data.
Meet Samture
Pick a slot, or send a message and we find a moment that suits you.