Using AI to transform the customer experience in banking

In my last blog, I shared some thoughts on how different forms of AI could help transform financial services, along with some challenges that needed to be overcome to ensure successful enterprise wide adoption.  As promised, this blog will focus on how AI can help improve the customer experience in banking; both directly, and more so indirectly behind the scenes.

When it comes to thinking about how AI can be applied to improve the customer experience in banking, many people will immediately jump to images of using “robots” to serve customers;  these could be actual robots (such as HSBC’s employment of “Pepper” in downtown Manhattan), or virtual agents (be it chatbots, robo-advisors etc) to service customer requests either on the phone or via chat tools.  Whilst HSBC saw some success with “Pepper”, it is safe to say that the more wide spread adoption of AI for customer service is focused on virtual agents.  Whilst virtual agents aren’t new (I remember helping a bank roll them out at least 10 years ago), their functionality has moved on massively from the heavily scripted solutions of a decade ago, to capabilities that utilise “Conversational AI” which help make the virtual agent interact with humans in a more human like way.  As well as the functionality moving on, so has the adoption by consumers – thanks to personal assistant technology such as Apple’s Siri, Amazon’s Alex and Ask Google, more and more consumers are happy to interact with services that they know are not backed by humans.  

Whilst employing virtual agents can absolutely help improve customer experience, AI can help transform so much more of the wider customer journey.  The foundation for great customer experience is to delight customers from the start – thinking of many of own interactions with banks, this is when we are first onboarded as a customer (be it taking out a loan with a new firm, moving your personal banking etc).  Any of us who have had to do this will know that the process hasn’t really changed over the years – there is an element of proving who we are (enabling the firm to satisfy that they have completed the “know your customer” [KYC] check), undertaking a suitability assessment of the fit of the product, maybe an element of credit scoring, or sanctions checking.  

AI can help in multiple areas across this whole process.  From a KYC perspective, Machine Learning can be deployed to help verify your identity by recognising government issued ID’s (Passports, ID Cards, Driving Licenses), and then matching them via facial biometrics to a picture or video of the customer (proving liveness etc).  Many banks are now deploying this approach to help enable them onboard customers digitally; UK challenger bank Monzo are a great example of how you can leverage AI to solve the KYC and AML compliance requirements.

For some products, such as mortgages or loans, a lot of documentation can still exist as part of the product origination process.  Some of this will be forms customers need to complete, as well as the usual plethora of “evidence” a customer needs to provide around earnings, bank statements etc.  AI – in this instance the use of Computer Vision, coupled with Natural Language Processing -  can be used to enable better processing of these documents by automatically analysing the data to make both decisions and also ensure compliance, as well as automatically populating systems of record.  This results in a more streamlined process from application to decision for the consumer.  However it is critical, when using AI to provide automatic decisioning, that the decision is explainable and is transparent.  

Through the life of a customer’s relationships with a bank, AI can be used to take the worry out of banking and provide more “value” to customers by providing insight into their banking habits.  Examples could include analysing the vast amount of transactional data banks hold to make personalised nudges to customers – as an example, if I’ve just purchased some plane tickets and I don’t have an active travel insurance policy (either directly with the bank, or known about due to transactions), it could provide a “nudge” to recommend that investigate taking one out.  Or a bank could look to provide near real-time personalised wealth management advice by leveraging advance predictive analytics to collect different data attributes about a customer, and then leverage algorithms to make many different predictions about the right investment strategy for a customer based on their goals.   Both these approaches also have the benefit that the bank can provide personalised experience at scale.   

Finally, AI can be used to help protect customers from bad actors.  As more consumers look to use less cash and increase the usage of other payments means (contactless cards, digital wallets etc), cyber fraud is on the rise.  Key to combatting this is the use of real-time fraud detection and fraud prevention solutions.  Historically, anti-fraud solutions were rules based, often relying on silos of data.  Due to the rules based nature of these solutions, they could only detect the “known” fraud scenarios.   These rules were also notorious for creating a high number of false positives – resulting in customer dissatisfaction with fraudulent transactions either not being detected, or legitimate transactions being blocked.  By adopting machine learning based real time fraud managements solutions, that are capable of performing sub-millisecond latency analysis of transactions, banks are able to understand the normal behaviour of each customer and use behavioural risk models that detect fraudulent transactions that are outside of the norm for that individual customer.   Ultimately accurately detecting more fraud and reducing the number of false positives; both of which results in happier customers. 

Many of the AI based solutions I’ve highlighted will both help positively transform the customer experience, as well as help to ensure that the bank remains in compliance with various industry regulations.  Whilst the use of AI can help further ensure regulatory compliance in areas I haven’t touched on, it must also be used in a manner that meets the regulatory standards.  More on this in my next blog.