CommBank deploys AI to spot emerging fraud patterns

CommBank has deployed an agentic AI system to detect emerging fraud and scam patterns in transaction and payment data. The deployment is part of the bank's AUD $1 billion annual spending on fraud, scams, cyber threats and financial crime.

The system is designed to identify new patterns in customer activity and generate rules to intercept suspicious transactions. It adds to a broader set of AI tools already used across CommBank's fraud controls.

Those systems monitor more than 80 million signals a day, spanning transactions, card payments, online payments and interactions across digital banking channels.

Australian banks have been increasing spending on fraud detection as scams and unauthorised transactions remain a persistent issue for retail and business customers. Lenders have also faced pressure to improve detection methods as criminals shift tactics more quickly and move across payment types and digital platforms.

CommBank has positioned the latest deployment as part of that broader effort to adapt its internal controls. The focus on rule generation suggests the new system is intended not only to detect unusual activity but also to shorten the time needed to turn emerging patterns into operational defences.

Fraud focus

Fraud systems in large banks typically rely on a mix of historical rules, behavioural analysis and machine learning models. One challenge is that scam and fraud patterns can change rapidly, leaving manual rule-writing processes struggling to keep pace.

By using an agentic AI approach, CommBank is seeking to automate more of that work within its fraud operations. In practice, the system is intended to identify previously unseen indicators in payment and transaction data and produce detection rules that staff can use to stop or review activity.

CommBank did not provide figures on the cost of the deployment or say how much of the AUD $1 billion annual commitment is allocated to AI. It also did not disclose how many fraud analysts or operational teams will use the system, or whether it has already led to measurable changes in scam losses or false positives.

Broader push

Financial institutions have been adopting generative and agent-based AI tools across a range of internal functions, from software development to customer support and risk monitoring. Fraud prevention has become one of the most closely watched use cases because banks process large volumes of real-time data and must make rapid decisions about whether to allow, delay or block payments.

That creates a difficult balance: banks need to stop criminal activity without disrupting legitimate customer transactions. Changes to fraud rules can directly affect payment flows and customer experience.

For that reason, many lenders have layered newer AI techniques onto established monitoring systems rather than replacing them outright. CommBank's description suggests the latest system is being integrated into its existing fraud framework rather than launched as a standalone platform.

Data volume

The scale of the monitoring operation points to the volume of information available for analysis. More than 80 million daily signals across transactions, cards, online payments, and digital banking interactions give fraud teams a broad dataset but also increase the need for automated tools to surface new threats quickly.

In recent years, scam detection has become more complex because payment fraud is often linked to broader social engineering attacks, including fake invoices, impersonation messages, and romance or investment scams. In many cases, the payment itself may appear legitimate in isolation, making contextual analysis and rapid pattern recognition more important.

The new AI system is intended to help detect those emerging patterns and create the rules needed to intercept them. The investment sits within CommBank's wider annual commitment to help safeguard customers from fraud, scams, cyber threats and financial crime.