Banking in the age of AI

Banking in the age of AI

Julia Krauwer, artificial intelligence expert at ABN Amro, discusses how advances in machine learning are increasingly being applied to an array of business-to-consumer interactions, including banking.

Like many people who were young in 1997, I had a Tamagotchi and it meant the world to me. Whenever it was hungry, I hurried to feed my digital pet. If it beeped about being lonely, I'd go give it attention. My parents thought it was a little silly to be that devoted to a toy, but in this day and age it's almost impossible to imagine life without interaction between man and machine. It’s a trend that’s here to stay.

Today's virtual sidekicks have come a long way, compared to my substitute pet of twenty years ago. Siri stands at the ready to search the whole web for you and even makes the occasional pre-programmed quip. Alexa understands what you're saying and uses a connection with your bank to transfer money as requested. Even your car can be linked to your digital personal assistant by Amazon, allowing you to remotely lock your vehicle and showing you your fuel levels at any time. Then there's Penny, a mobile app with an avatar by the same name which can maintain a natural dialogue about your income and expenses. Nice little detail: Penny uses emoticons just like a real person.

As basic or highly advanced as they come, digital assistants are being introduced to many areas of life. Most of them don't pick up on our emotional signals just yet. The MoodBox is an exception; it selects and plays music for you based on the emotion in your voice, and tops it off with sympathetic reactions. Amazon is developing an update for Alexa which will enable it to notice if you sound bothered, and adjust its answers accordingly. Not only are these tools becoming incrementally clever with each update and new generation, they're also turning more and more human in the process. Even artistic creativity doesn't seem to be a human-only trait anymore, now that the first computer-generated tunes, paintings, and novels are being released.

We have not yet developed a computer that can simulate all cognitive, emotional, and creative abilities that humans possess. However, artificial intelligence is already seeing interesting applications in various areas of life. In the financial world, we divide them into two types of virtual assistants: those that help bank employees do their work more efficiently, and those that directly interact with clients.
Smarter with each conversation

We've already implemented many features that fit into the latter category. A chatbot on the SwedBank website answers all visitor questions, and challenger bank Atom uses an app with a virtual agent who gets smarter with each conversation. And recently, during Startup Friday, we showcased a prototype of a new chatbot that tells you, after you answer a set of questions, what your maximum mortgage is. These assistants don't always have to come equipped with a face and a personality. Some digital assistants are simply smart apps, such as ABN AMRO's Grip which offers insight into your income and expenses.

The bank employee, too, benefits from the help of digital assistants. Imagine a dashboard for the relationship manager with a one-page overview, listing clients that need advice and their related financial needs. Or a virtual robot that picks up the monotonous and time-consuming tasks. Another application is a fraud detection system that traces subtle changes in transaction patterns. Digital assistants can run risk analyses that eliminate gut feeling, and rely purely on advanced data analysis.

No matter which way you look at it, Artificial Intelligence can help companies provide customers with even better service. Directly by means of chatbots and smart apps, or indirectly by empowering employees and freeing up their time and headspace. When it comes to services we've already automated, we provide our clients with a personal touch in virtual form. Other services, with human contact at their core, receive a boost in the amount of time and personal attention we give the client.

Human contact and personal attention - I'm proud of how we handle it. Just like how I used to share my day-to-day worries with my girlfriends and family when I was younger, I still maintain that preference for speaking with another person. That human contact will continue playing a pronounced role in dealings between ABN AMRO and its clients. Simultaneously, interaction between man and machine is becoming more and more commonplace. At the bank we keep tabs on these developments so we can continue providing a healthy balance in the future. We're living in an era where people entrust chatbots with their secrets, and answer virtual psychologists' questions until they get to the bottom of things. And although my Tamagotchi simply ended up in a box after its demise, Japanese AIBO dogs are given a respectable funeral.

Will we eventually see today's virtual assistant develop into your future virtual BFF? I’ll ask my Tamagotchi what it thinks. Or my girlfriends...

Comments: (1)

Juergen Rahmel
Juergen Rahmel - IETC Information Engineering Ltd - Hong Kong 12 October, 2016, 04:53Be the first to give this comment the thumbs up 0 likes

Interesting and entertaining write-up of the high-level, sunny-side-up features of AI in the Banking scenario. Thanks for this.

I suggest that in parallel we also start publicly looking ‘under the hood’ of the approaches and promises. AI is not a singularity, AI is a large variety of capabilities, types of knowledge representations, learning mechanisms. Each of them has properties, features, risks and issues to it.  

Using one simple (AI-free) example from the text: Being able to lock my car remotely is maybe a cool function, but it means at the same time I provide the infrastructure for someone else to remotely open it. I personally lock my car from nearby, when I leave it. I’d never want this to be a truly remote feature. I do not have a problem that asks for this solution. So I’d better avoid the costs and risks that come with it.

Same is applicable in the banking and insurance context, when deploying AI solutions. We must come from the problem side, look into solution options (an many times, simple statistics is the best solution) and consider their applicability as well as their risk profile (e.g. false positives, false negatives and their consequences in the given context). This is valid for both, the customers and the institutions.

I believe this space needs a lot of education of customers, businesses and IT departments in order to shape and manage the expectations – and reduce the frustrations that are upon us in the brave new smart world we are heading towards.