Subscribe to stay notified
I Understand We use cookies to help us in different ways, such as with our marketing and to make our website better. See our Cookie Notice for more information.
If you continue to use our website, we’ll assume you’re OK with us using cookies.

Media

Smarter, Better, Faster, Stronger: How AI is revolutionising KYC

Matthew Warner | December 2019

Whilst we all acknowledge the necessity of KYC, there are very few of us who actually enjoy the process. The time it takes, the cost, the inconvenience; all these factors rack up the pain of keeping companies and users safe from malicious actors. As digital economies grow and markets push for fast, global services, the volume of data not only creates a wealth of opportunity for criminals to exploit, but also causes an issue in the sheer amount of data that needs to be processed and analysed for verification.

Traditionally, checks were carried out by humans, comparing documents and data before approving or rejecting individual cases. This meant a slow and labour-intensive process, with both businesses and customers frustrated at the delay. Indeed, recent studies show that users prefer quick and simple solutions to those that are more thorough and secure – they are willing to sacrifice safety for convenience. Onboarding new customers was therefore a real problem, both in cost and time, with some banks estimated to spend potentially weeks and up to 10% of their revenue on KYC and due diligence, with  the demand and traditional resources required to meet it only rising. Fortunately, technology has once again provided a solution. 

The use of Artificial Intelligence (AI) has potential in many areas and industries, but in the past few years, it is increasingly being looked to in order to solve the issue of growing KYC demand. Going through its strengths, it’s not hard to see why it matches up well. KYC requires a large amount of analysis and processing, something that machines excel at. AI has the capability to handle volumes of data far in excess of what any human could ever hope to manage, and with a greater accuracy – not being distracted or indecisive and able to find inconsistencies or errors in both large amounts of data and minute detail; AI has consistently been shown in studies to outperform humans in repetitive tasks such as verification. Unlike a human verifier – the traditional KYC analysts – a computer can run 24/7, with no need for breaks or rest, a vital element in providing KYC that doesn’t inconvenience customers. 

Not only do machines have working and processing advantages over humans, but they also don’t suffer from any preconceptions or biases that could influence their judgement. An AI solution also has the ability to translate languages almost instantly and can learn and be kept up to date with the most recent regulations and standards without confusion. In addition, using the data, the AI can compile risk profiles with a high level of accuracy and categorise them, produce audit trail reports, analyse metadata, automatically flag issues as they arise, and anything else it can be programmed for. 

One final advantage an AI-based system has is in its perception. People may feel uncomfortable about other humans having access to and analysing their personal data, but a machine that has no emotions and does not judge is unlikely to be an issue for anyone. 

Despite all its advantages, it is important to note that AI is not a panacea for KYC. Even with its ability to analyse and sort information and documentation, there are still fringe cases which need human intervention to resolve; a mixture of AI with human intervention where needed has been suggested for most KYC solutions. There is also a potential issue with security – as with any instance of having large amounts of data – it acts  as a honeypot for attackers. Efforts are being made to solve privacy issues for good, with methods such as zero knowledge proofs and homomorphic encryption being developed, and blockchain itself has the potential to bring security and immutability to the solutions. Regardless, it will take time and effort for these various elements to come to fruition and possible further developments to improve AIs functionality even further. Until then, our work is never over … but with AI it can be eased.