The application of artificial intelligence (AI) to solve human problems is not a new technique. AI has been around for decades in some form or another. In the early 1960s, Stanford Research Institute (SRI) created ‘Shakey the Robot’, an AI-powered robot. But it wasn't until the 1990s that AI, as used in commercial computing tasks, really began to find a home with the development of ‘Intelligent Agents’.
Now, in the 2020s, the financial sector is fully embracing the use of AI. This is being reflected in the market for AI-enabled solutions. The global market for AI in Fintech was estimated to be around $6.67 billion in 2019. By 2025, this is predicted to grow by 23.37% to $22.60 billion.
AI is being used across the financial sector use case spectrum, with challenger banks forging ahead with the technology. But AI isn’t just for the challengers, banking, across the board, is looking to improve customer experiences, efficiencies, and reduce fraud, using AI.
Four Use Cases Using AI in Finance and Banking
Artificial intelligence (AI) is a swiss army knife with subsets of the methodology, including machine learning (ML), deep learning (DL), Natural Language Processing (NLP), etc. each having specific applications in banking. For example, Machine learning, a subset of AI, is based on algorithms, a set of rules that use data to learn how to complete a given task. The data is used to train the ML algorithm without the need for any hard-coded programming. This makes software solutions based on ML highly flexible and versatile. Software based on ML uses large amounts of data, looking for patterns and spotting outliers, e.g., anomalies in data and behavior. It can then use these to process many tasks from anti-fraud checks to tailoring customer experiences.
AI and the subset, ML, are being increasingly used to solve some of banking’s most complex issues. Here are four such use cases that are seeing good results in the application of machine learning:
Banking, payments, and fraud detection
A 2020 report from analyst firm Juniper Research, predicts that by 2024 online payment fraud losses will reach $200 billion. The report also highlights that spending on AI software (in the form of machine learning) for fraud detection and prevention, will reach $10 billion in 2024, an increase of 15% on 2020 figures.
The reason for these massive losses is in large part due to fraudulent card-not-present and identity related theft. AI, or more accurately, machine learning (ML) provides a mechanism to improve the capture of fraud events across the entire lifecycle of payments. ML algorithms learn patterns and trends using payment data. ML anti-fraud solutions can then use these patterns to detect anomalies, these differences in expected behaviors of individuals or systems, potentially pointing to a fraudulent event.
A recent EastNets survey looked at how ML helped organizations deal with fraud on SWIFT networks. With regard to SWIFT payment fraud in banking, 85% of leaders in the space used computer-based user behavior analytics (based on ML techniques) to reduce fraud. DownloadSWIFT Cyber Fraud Survey Report
The targeting of payment lifecycle events is a behavior that is prevalent in the world of cybercrime. Payment transactions have many points in the process where fraud can occur. Sanction screening, including checks for individuals on PEP (politically exposed person) lists, is an example where ML can be applied. The ability to prevent fraudulent transactions is also mandated as part of anti-money laundering (AML) check requirements that meet a variety of regulations in banking, e.g., The Bank Secrecy Act in the USA.
Sanction screening is a technique used to spot potential cyber-fraud occurring during certain transactions. However, traditional Transaction Monitoring Systems (TMS) have very high false positive alerts during AML (anti-money laundering) checks, often as high as 90%+. This results in poorly optimized systems that often fail, annoying legitimate customers and costing the bank in terms of resources. Machine Learning based advanced analytics, improves optimization of transaction monitoring, reducing false positives, improving accuracy, and reducing customer friction.
Using AI and ML for smart analysis of data helps to maintain payment speed and improve accuracy in fraud detection, whilst preventing cybercrime.
Customer experience and chatbots
A great customer experience (CX) is a driving force in the finance industry. The “2020 Digital Marketing Trends Report” from Adobe identified CX as the single most important opportunity for an organization in 2020.
Cybersecurity and compliance
The finance sector has come under increasing threat from cyber-attacks in recent years. Cyber-attacks against major brands such as Capital One have made international headline news. A report from the Federal Reserve Bank of New York points out that cyber-attacks are increasingly able to leverage the interconnectivity of banks; a cyber-attack having the potential to impact up to 38% of the banking network.
A report from Boston Consulting Group found that financial services are 300 times more likely to be the target of a cyber-attack than other sectors. The general area of cybersecurity threats against financial institutions can benefit from the application of an AI-based cybersecurity approach. As noted earlier, payment fraud uses ML systems to detect fraudulent transactions for AML purposes. But other areas such as cybersecurity threat monitoring and the aggregation of cybersecurity event data are enhanced by ML techniques. These data is used to augment traditional monitoring and detection of cyber-threats, providing intelligence to security analysts in the fight against cybercrime.
One of the related areas that ML/AI based services can provide in banking is to augment and facilitate compliance needs. A report from Computer Services Inc. shows that 75% of banks expect to spend around 20% of banking budgets on regulatory compliance. Regulations, like the New York 504 “Final Rule” (NYSDFS part 504) are being updated with modern smart technologies in mind. In the area of AML regulations, again, AI-based anti-fraud solutions can help to meet these stringent regulatory requirements. The use of artificial intelligence and machine learning to adapt to sophisticated cyber-threats helps a financial organization to meet data protection and AML regulations.
Process automation (RPA)
The use of Robotic Process Automation (RPA) is one of the driving forces behind what Capgemini describes as “New Age Banking”. RPA uses AI to improve operational efficiencies. AI techniques such as Computer Vision, ML, Natural Language Processing (NLP), and deep learning (DL) can be used with complex, document-led, processes to automate and simplify banking processes. These processes including account transaction events, underwriting, and billing, are all being enhanced using RPA.
The Human Hand in AI
A recent report by the Economist Intelligence Unit (EIU) stated that AI will “separate the winning banks from the losers”. However, the report caveats this by pointing out the importance of addressing any bias in the use of AI.No thanks thank you The report writers stress that human monitoring and review of AI-based decisions are needed to balance any fairness issues that could arise in AI-based decisions.
This co-dependence between AI solutions in banking and human operators is likely to be one that will create a symbiotically efficient service in the financial sector, moving forward.
Banking on an AI-Enabled Future
It is likely that a combination of compliance needs, anti-fraud urgency, and seamless customer experiences, will drive further adoption of AI-enabled banking solutions. Banking is at a juncture. A perfect storm is brewing. A storm that takes into account:
- Customer expectations of digital services that are seamless and speedy
- Back office efficiency enhancement
- Stringent AML regulations to deal with
- A highly sophisticated payment security threat landscape to mitigate
As banking further evolves into a data-driven digital business that offers fully online financial processes, the sector must turn to smart technologies. To maintain exceptional customer experiences, banks can use friction reducing machine learning enabled platforms, such as AML solutions.
The ‘New Age Bank’ can expect to be smarter, more agile, and customer-friendly, whilst maintaining secure systems.