Bank of Ireland prevented €9.7 million in customer losses in 2025 after deploying artificial intelligence to assess card transactions for potential fraud, underscoring how machine learning is being embedded into core banking controls. The system analysed one billion card transactions over the year, identifying suspicious activity in real time.
According to the bank, the most prevalent fraud affecting its customers was investment scams, followed by re-direction and romance scams. While card fraud accounted for a higher volume of cases, individual losses tended to be lower in value. Approximately 80 per cent of card fraud cases occurred through online transactions, with the remainder largely linked to social engineering tactics such as smishing and phishing. The scale of attempted fraud was particularly evident during the holiday period, when the bank’s fraud prevention team received more than 10,000 calls between 23 and 29 December 2025.
The lender’s fraud detection framework is built on a machine learning scoring model that incorporates behavioural analytics and anomaly detection. Agentic AI is also used to power workflow management within fraud operations, enabling faster triage and response. Alongside its technology investment, Bank of Ireland maintains a fraud and financial crime department comprising 225 employees and has established an AI academy that has upskilled more than 1,000 staff, indicating a parallel focus on internal capability development.
The initiative illustrates how AI and machine learning are being applied within payment processing environments to mitigate financial crime exposure at scale. The operational demands highlighted by seasonal spikes in fraud reports point to the continuing need for adaptive systems capable of responding to evolving scam patterns, particularly as digital transaction volumes remain high and fraud tactics grow more sophisticated.
