I recently reviewed one of the best books written on AI to date, AI Super-Powers, where a brief history of AI is explored across the four waves of evolution experienced over the last few decades. From chess-playing, payments, making friendships, medical diagnoses, and help with investing money, AI continues to play a much broader role across many market segments. In a 2019 study by datafloq, 6 sectors were identified where the adoption of AI had been the most prolific, namely;

• Healthcare – where efforts have focused around having access to remote digital advice and medical consulting, assisted surgery, and management of records
• Education – here development has primarily focused on AI-enabled scheduling for parent-teacher check-ins 
• Marketing – as more and more consumers use online channels to learn about products, the more significant the role of AI has become in targeting the right offers to the most receptive audience
• Retail & eCommerce – where the first wave AI is still very much in play, with recommendations forming a large part of driving sales e.g., others like you also found this useful
• Financial Services – banks were rapid to adopt automation several years ago to make more homogenous and repetitive tasks more cost-effective to execute. Now the next wave of intelligent automation has become more popular as a mechanism to improve operating model economics

If we switch our focus to Financial Services, AI adoption is outstripping any other non-manufacturing sector, with players using the technology to solve real business issues. Here we examine the top 6 use cases across banking where AI is making a difference. Each area has been selected based on the size of the issue, cost of executing the process, the ability to positively impact the customer experience, and lastly, the ability to drive significant cost savings to a bank or financial services provider 

KYC/KYB/AML and Customer Onboarding – This process, which is key to onboarding customers for any regulated banking product, still seems a challenge for incumbent banks. The process is still document-hungry, and the checks undertaken to determine a person is who they say they are, have not changed for several years. AI can help streamline this process by allowing for verification techniques like facial recognition, or automation of document uploading, using OCR to pre-populate application data as well as proving verification of identity

Anti-Money Laundering (AML) - Transaction monitoring has always been a challenge for full-service banks. The primary issue is false positives, where genuine transactions are flagged as suspicious. Outside of the inconvenience for the customers, these transactions are costly to process. AI can play a vital role in addressing this long-standing pain point by helping banks move from rule-based analysis to more risk-based assessments 

Servicing queries – Servicing is such a critical part of any banking experience, but due to the under-investment in communication channels, voice servicing has become expensive. The cost of servicing becomes more uneconomic when a simple transaction such as balance checking or getting information on transactions can be undertaken in a digital self-service environment. AI utilised appropriately can help to drive more efficiency when dealing with customer queries, by using a combination of chat-bots, dynamic FAQs and Robo-advisory

Collections – The post-pandemic climate is expected to be more delinquency-prone than we have seen in the previous few years. This then brings about the challenge of Collections after customers default on loan payments or other products. Using AI, setting, and executing a first, second, and third line of defence strategy becomes less labour intensive. Here data can be used to determine the best channel, time of day, and communication style that will yield the best opportunity to collect overdue payments

Fraud – Another expensive area for banks is Fraud, especially as more and more interactions are taking place remotely, enabling newer forms of Fraud to emerge. AI can help to prevent, as well as identify Fraud, through analysing customer behaviour in real-time to determine what activity looks out of kilter with normal behaviour. The accuracy, and hence ability to avoid false-positives, can also be improved over time by looking at more of the second and third-party data available to not only support in assessing valid transactions, but also make identity verification stronger through biometric-based techniques 

Underwriting – For any onboarding process, where it is for a current account, credit card, loan, or mortgage, underwriting is usually the longest part of the process. In some cases, we have seen onboarding take as long as 12 days to complete in the UK because the underwriting process has been manual. This timeline is driven by the need to have someone physically eyeball information, create a risk report, and then assess all that information alongside a scorecard, and then decide to “accept” or “reject”. In this regard, if we think about document management, information processing, and generating decisions as largely rule-based, then machine learning can make a significant difference to the time it takes to structure the data and then support decision making. These techniques allow onboarding timescales to be reduced drastically, as well as allowing for real-time decision making, caters for a much better customer experience, as well as a low-cost one

In summary, if re-designed and automated intelligently, all these areas have a real ability to move the dial on improving both customer experience and reduce the operating cost. The savings here are significant, i.e., savings over 30% (based on my experience of working with banks across the globe)... which, in this economic climate, would be a sensible thing to consider as part of any banking strategy.