Neobanks, artificial intelligence and KYC

Artificial intelligence is changing the neobanking landscape, and machine learning is poised to improve your identity verification process, boost your conversion rate and accelerate your ROI.

In traditional banking, a financial institution would try to prevent financial crimes and illegal activities, such as terrorist financing, by implementing KYC-compliant practices. These KYC compliance measures would include physically checking a potential customer before they have access to day-to-day services, from opening bank accounts or a current account to purchasing new debit or credit cards. In financial services, it is still necessary to verify the identity of potential customers, but with the growth of digital banking globally, operators are looking for more innovative solutions to complete enhanced due diligence (EDD), customer due diligence (CDD) and other compliance measures.

The application of artificial intelligence (AI) in the form of machine learning promises to be transformative for many sectors, and neobanking is no exception. By analysing huge volumes of digital data, automated decision engines can recognise patterns and progressively improve the speed and accuracy of processes. This type of machine learning is already being used with great success by neobanks, both to improve fraud prevention and to provide a better user experience.

Neobanks are increasingly using identity verification (IDV) for the know-your-customer (KYC) process required by regulators to prevent money laundering. The best IDV systems compare live biometric data from a selfie with stored biometric data to authenticate customer sessions. By incorporating machine learning algorithms programmed to identify patterns into this type of automated IDV solution, the collected data can be used to adjust and improve the accuracy of decision making over time.

Because machine learning makes predictions based on previous experience, novelty is a potential blind spot for this technology. Therefore, the larger and more current the dataset, the lower the risk of an automated decision engine making a mistake. For identity verification, for example, access to a dataset that includes numerous biometrics for a wide range of ethnicities and ages is important to minimise potential biases. In addition, cross-matching verification sessions based on device, network, and customer behavior can provide a richer and more varied set of data that a decision engine uses to learn and improve its decision-making.