How bank combines AI and human intervention to stop scams

With a few keystrokes, Dr Chen (not his real name) transferred a total of half a million dollars – savings he had toiled for decades to accumulate – to another account.

He had locked himself in his clinic’s office for five hours while staying on the line with callers whom he believed were government officials. They told him they were investigating money laundering charges against him and that he had to make multiple fund transfers from his DBS bank account in order to prove his innocence.

What he did not realise was that these people were not government officials, but scammers.

Fortunately, his transfers did not go through.

Less than a minute after Dr Chen clicked “Transfer”, DBS’ systems flagged the suspicious transactions. An alert flashed on the screen of Ms Chloe Lim, an analyst at the bank’s anti-scam team, prompting her to contact Dr Chen immediately.

She could not reach Dr Chen, but managed to reach his wife who was listed as a joint account holder. Ms Lim explained to the couple they were likely victims of a scam and obtained their consent to halt the transfer process.

With Dr Chen’s approval, Ms Lim quickly contacted the bank receiving the transfers and had the funds frozen before they could be withdrawn.

This incident illustrates one of the many alerts that the DBS’ anti-scam team receives every day. The alerts represent only a fraction of potential threats, as the bank’s artificial intelligence systems quietly intercept and block countless suspicious transactions before they even reach analysts like Ms Lim.

DBS bank’s invisible shield against scams

DBS’ ability to identify and prevent such high-risk payments is the result of continual investments in its surveillance capabilities.

The bank’s fraud detection and prevention systems analyse transactions in real time, using artificial intelligence alongside traditional rules-based checks to identify unusual transaction activity.

But as scams become more sophisticated, detection alone is no longer enough. Increasingly, scammers are manipulating victims into authorising transactions themselves. To address this, DBS has in place a multi-layered scam defence which includes “helpful friction” in the form of cognitive breaks built into higher-risk journeys, to slow customers down at critical moments and prompt them to pause and reconsider before proceeding.

These measures range from alerts and additional verification steps, to transaction holds and direct outreach by the bank’s anti-scam team, depending on the level of risk detected.

“Transactions continue to be processed normally with no noticeable delays,” says Ms Precious Riola, senior vice president, Group Fraud Detection Analytics, DBS.

The goal, shares Ms Riola, is to find that delicate balance between protecting customers while making the banking experience feel seamless. This is done through a “tiered intervention” embedded into the surveillance system, where security measures are calibrated based on the perceived risk of each transaction.

For example, if the system detects something slightly unusual with a low risk score, it would send an alert to customers, prompting them to contact the bank for assistance if needed.

But if it detects high-risk activity, such as unusually large transfers or certain transaction patterns that suggest fraudulent activities, the system takes stronger action, temporarily holding the transaction while the anti-scam team reaches out to verify.

In the event of unusually rapid outflows from a bank account, the transactions will also be placed on hold for 24 hours, with immediate notifications sent to account holders. This measure is part of a Shared Responsibility Framework introduced by the Monetary Authority of Singapore in June this year to give both customers and banks time to verify the legitimacy of transactions before funds are released.

Human judgement still essential

Even with advanced surveillance, human judgement remains a critical part of the bank’s scam defence.

Where banks once primarily fought unauthorised fraud, they now face a more complex threat: scammers who convince victims to willingly transfer their money.

Says Mr Alan Teng, head of DBS’ anti-scam team: “Technology can spot suspicious patterns, but it’s often human intervention that breaks the scammer’s control.”

Dr Chen’s case illustrates why. While AI detected the unusual transfers within milliseconds, it took Ms Lim’s careful intervention and explanation to help Dr Chen recognise the deception.

“That human connection is so important,” Ms Lim explains. “Scammers take advantage of people’s fear and urgency, making victims feel isolated and pressured to act quickly. Victims are coerced into providing personal details or transferring funds for ‘safekeeping’.”

She recalls cases where scammers used video calls, complete with fake uniforms and forged documents, to appear legitimate. “They tell victims to keep quiet about supposed investigations, cutting them off from anyone who might help them see through the deception.

“But when we reach out, we break that isolation,” she adds. “By asking gentle questions and offering reassurance, we help customers step back and recognise the warning signs.”

Frontline vigilance is a key layer of DBS’ scam defence. The bank’s anti-scam team and frontline staff undergo regular training to recognise behavioural cues – such as defensiveness, inconsistencies or unusual urgency – and to intervene with empathy and care.

With the help of technology, Ms Lim has helped stop countless fraudulent transactions, including a single case where she prevented $200,000 in losses. But the impact of her work extends beyond numbers. “Behind every alert is someone’s life savings or family business,” she says. “That’s what gives me satisfaction in ending my day knowing I’ve saved someone from losing their money.”