Professor Jon Crowcroft, iKVA
Money laundering costs the UK economy £24bn each year, according to The National Crime Agency, whilst the UN estimates that up to $2tn is moved illegally each year via large, multinational banks. The surge in financial crime, triggered by the pandemic, has huge implications for the finance industry, as areas such as Anti-Money Laundering (AML) come under heightened pressure to prevent fraudulent activity.
The pandemic has accelerated both the adoption of digital technologies and organisational readiness, with a third of financial institutions hastening the use of Artificial Intelligence (AI) and Machine Learning (ML) in their AML strategies to fight the growing problem. Firms are thinking strategically about future proofing their technical abilities, and AI technology can help them utilise existing data to pinpoint trends and identify risks, while conserving manpower by breaking down data siloes and enabling data discovery.
Preventing fraud
Fraud, whether this is money laundering, facility takeover or application fraud, represents a real hazard for the financial sector. The analysis of substantial, siloed datasets using traditional methods of fraud detection, requires significant investment of human time and labour to scrutinise datasets and generate accurate pattern predictions.
AI technology can accelerate the examination of vast amounts of information and identify suspicious patterns and behaviours in a fraction of the time that humans can. Using large data sets, that humans would be unable to process, also enables AI to learn fraudulent data patterns and so improve detection accuracy and frequency over time.
One branch of AI, Federated Machine Learning (FML), is currently being used by the Financial Conduct Authority (FCA) to monitor fraud on a global scale. FML uses the latest AI techniques to understand the patterns of fraudulent transactions which can then be deployed internationally on decentralised edge devices hosted in individual countries. Being deployed locally enables the FCA to avoid contravening local regulations about data transfer, while benefiting from the knowledge gained from the data models. The application of AI not only removes the expense of human interaction, it increases the ability of an organisation to detect fraud, since AI typically detects patterns more accurately than humans. In 2019, the FCA used AI technology to analyse 10,000 data points across 16 regulatory returns, including their Financial Crime Data, to identify potential outliers and produce analytics to highlight risk profiles.
AI technology has a similar application in financial trading. Analysis of trading can identify patterns and combinations of patterns, that are likely to indicate suspicious activity that would indicate investment fraud. Nasdaq has introduced a deep learning system to monitor for fraud in the more than 17.5 million trades done daily on one of the world’s largest stock exchanges.
U.S. Bank is using AI technology to unlock and analyse relevant data on customers, via deep learning, and identify potential transactions that may indicate money laundering. This has doubled the output, compared with the prior systems’ traditional capabilities, clearly demonstrating the efficiency of AI software in identifying anomalous activity.
Finally, Visa’s artificial intelligence (AI)-driven security helped financial institutions prevent more than AU$354 million in fraud from impacting Australian businesses between May 2020 and April 2021. The algorithm assesses more than 500 risk attributes in roughly a millisecond to produce a score of every transaction’s predicted fraud probability. This is then used to identify patterns and alert financial institutions to potentially fraudulent activity – before it even takes place.
Reducing risk
Accurately evaluating the financial risk of loan underwriting and credit applications, relies on multiple datapoints from a multitude of sources, which often have different search interfaces, requiring many and repeated searches to find information.
AI technology can overcome this by bringing all of these sources of information, such as bank statements, pay slips, tax filings, mortgage forms and invoices, together in one place, indexing and segmenting the data, and allowing this to be discoverable. This provides for better and more accurate results and means that decision sources can be quickly and easily identified and evaluated, helping to reduce business risk. Companies that estimated their profit from Big Data analysis have reported an average increase in revenue by 8% with a 10% reduction in costs.
Lost in translation
The finance industry generates huge volumes of data and, with a wealth of data sources worldwide, accessing filings, transcripts, research and news to discover changes and trends in financial markets in an efficient manner can be difficult – especially when the information is in different languages. For multinational corporations, operating on a global scale, time and money must be spent on translation services to understand the results.
However, Natural Language Processing (NLP) combined with Neural Networks and deep learning models, enable computers to understand the full meaning of information including sentiment and intent without the need for translation. So, advanced language agnostic AI tools, such as those developed by iKVA, enable finance operatives and regulators to discover relevant information, regardless of the source language.
Conclusion
The benefits that AI technology can bring to the financial industry are vast; saving time, reducing costs, improving compliance and reducing business risk. Many leading financial companies are now deploying AI on a global scale and starting to reap significant benefits, but AI is still in its infancy in the sector compared both to other industries and the potential of this technology. Appropriate utilisation of AI technologies is becoming paramount for the financial sector to reduce ever increasing fraudulent activity and to remain competitive.
