Summary

• Bank of Ireland says more than four in five card chargeback cases close within three days, while AI-assisted systems examine approximately 2.9 million transactions daily for potential fraud.
• Its Benefits Finder has been used 31,000 times and identified €55 million in potential tax refunds; that amount does not represent confirmed payments to customers.
• Nearly 7,000 employees have completed Microsoft Copilot training, with users reporting average daily time savings of around 30 minutes. The bank has not disclosed a consolidated return on its AI investment.

Bank of Ireland has published operational figures from its growing use of artificial intelligence, with applications processing card disputes, examining millions of transactions for fraud and helping employees complete administrative work.

The bank says more than four in five chargeback cases now close within three days, while its AI-assisted systems assess approximately 2.9 million customer transactions every day for indications of fraud. Almost 7,000 employees have completed Microsoft Copilot training, and users are reporting average time savings of about 30 minutes daily.

The figures show AI operating within established banking processes rather than remaining confined to demonstrations. They do not, however, provide a complete financial assessment of the programme, and the bank has not disclosed comparable performance baselines or independent measurements for every application.

Bank of Ireland coordinates deployment through a central AI Centre of Excellence, which identifies opportunities, establishes priorities and oversees the use of technology within what the bank describes as a secure, governed operating environment.

Card disputes provide an operational measure

Chargebacks allow cardholders to challenge qualifying transactions, including cases involving unauthorised payments or problems with goods and services purchased.

Resolving a dispute can involve examining transaction records, obtaining information from customers and merchants, applying payment scheme requirements and coordinating activity between financial institutions.

Bank of Ireland says AI is supporting that work and reports that more than 80% of cases are now closed within three days. Faster processing can reduce uncertainty for customers, particularly where the underlying transaction involves money they expected to remain available.

However, the bank has not published a comparable historical resolution rate alongside the latest result. It is therefore not possible to calculate the precise improvement attributable to AI or establish how much of the current performance reflects other changes in staffing, procedures and supporting technology.

The distinction is important when financial institutions report progress through operational milestones. Closing a large proportion of cases within a specified interval is a service outcome; determining what caused that result requires additional evidence.

Financial institutions also need to monitor accuracy and customer treatment alongside speed. Resolving disputes more quickly would provide limited benefit if the process introduced incorrect decisions or failed to identify cases requiring closer examination.

Fraud screening operates at transaction scale

Bank of Ireland’s AI-assisted fraud technology examines approximately 2.9 million customer transactions every day, throughout the year.

Fraud monitoring systems assess activity for patterns associated with scams, compromised accounts or unusual payment behaviour. Machine learning can assist by identifying relationships and anomalies across volumes of transactions that would be impractical for employees to inspect individually.

The effectiveness of such systems depends on their ability to detect genuinely suspicious activity while avoiding excessive false alarms. A legitimate customer making an unusual payment may resemble a fraudulent transaction when examined without sufficient context.

Overly sensitive controls can delay payments or create unnecessary customer contact, while inadequate detection may allow suspicious activity to proceed.

The 2.9 million daily figure measures transactions examined, not confirmed cases of fraud prevented. The bank has not included detection accuracy, false-positive rates or the financial value of prevented fraud in its latest announcement.

Bank of Ireland has previously discussed technology used to combat payment fraud, but the new report concentrates on the operating scale of its AI programme across several areas of the business.

Assessment of the programme’s fraud effectiveness would require more detailed evidence about outcomes, investigation costs and the relationship between automated recommendations and decisions taken by specialist teams.

Financial wellbeing tools identify possible refunds

Customer-facing applications form another part of the programme, including Benefits Finder, which helps people identify tax credits and reliefs for which they may be eligible.

Bank of Ireland says the tool has been used 31,000 times and identified €55 million in potential tax refunds.

The identified value is not equivalent to money transferred to customers. Eligibility must be established under the relevant tax rules, and customers may need to complete additional steps before any refund is approved or paid.

The usage figure also describes the number of times the tool has been used, rather than the number of unique people who have received payments or completed successful claims.

The application illustrates how banks can use automation to extend services beyond payments, savings and lending. Financial institutions already have established customer relationships through which tools offering practical assistance can be distributed.

Whether those tools deliver sustained financial benefits depends on the quality of the underlying guidance, the accuracy of the information used and the proportion of customers who act successfully on the results.

The bank separately reports delivering 127 million AI-powered spending insights during the previous year.

These insights may help customers understand patterns in their financial activity, although delivery volume does not demonstrate that the same number of individuals engaged with them or changed their behaviour as a result.

Contact centres use AI to reduce transfers

Bank of Ireland also reports a reduction of more than 40% in call transfers associated with AI use in its contact centres.

Routing enquiries effectively can reduce the number of times a customer must repeat information or move between support teams before receiving assistance.

AI tools may help identify the purpose of an enquiry and connect a customer with the most relevant support process, although performance depends on the information available and the system’s ability to recognise when specialist assistance is necessary.

The bank has not published corresponding results for overall call handling time, customer satisfaction or the proportion of enquiries resolved during the first contact.

Those measures would help establish whether fewer transfers also produce better service outcomes. A contact may remain within one team while still taking too long or failing to resolve the underlying problem.

Financial institutions must also provide suitable support for customers whose circumstances cannot be understood through routine automated processes, including people who may need additional assistance.

Copilot training reaches thousands of colleagues

Almost 7,000 Bank of Ireland employees have completed training in Microsoft Copilot as part of a wider rollout.

The programme includes ordinary workplace tasks such as research, summarising information and preparing routine materials, with specialist training also available through University College Dublin and Cambridge Spark.

The bank says employees using Copilot save an average of approximately 30 minutes per day.

That figure has not been accompanied by a published independent evaluation explaining the sample, measurement method or comparable results from employees not using the technology.

Time saved on individual tasks should also be separated from actual financial returns. An employee may redirect recovered time towards customer support, analysis or other activities without creating a measurable reduction in operating expenditure.

Conversely, licensing, integration, information security and output verification can create additional costs, particularly where generated documents or conclusions require human review before use.

These considerations are especially relevant in banking, where systems may process commercially sensitive material and information about customers.

The Centre of Excellence is intended to coordinate the bank’s approach, providing a common point for evaluating uses and governing deployments. Central oversight can help limit fragmented adoption, although its existence alone is not evidence that every application has been independently validated.

Bank of Ireland’s latest figures show AI assisting several operating processes at substantial scale, from routine card disputes to fraud monitoring and employee administration.

The next assessment of progress will depend on more comprehensive evidence about the costs of those deployments, the quality of their outcomes and whether improvements remain reliable as the bank extends the technology into additional functions.