Historically, the banking sector has been a kind of ocean liner that has struggled to change direction, clinging to legacy technology and a mindset from earlier times. In terms of technological innovation, banks have been overtaken by the fintech ecosystem in terms of agility. However, as banks try to catch up, it’s important to take a hard look at the promise of artificial intelligence and what it means for the banking industry.
Manuel Morales wears a variety of hats in machine learning and banking. Since 2018, he has been General Director of the FinML Network. He is currently an Associate Professor in the Department of Mathematics and Statistics at the University of Montreal specialising in Financial and Actuarial Mathematics, specifically the application of Machine Learning in banking and in responsible investment.
He recently participated in the Artificial Intelligence transformation initiative of the National Bank of Canada. As the Chief AI Scientist, he led the scientific efforts of the bank’s strategy to leverage AI technologies across all verticals. He had the opportunity to work on a wide variety of projects from wealth management to retail banking applications, and also led the first works towards laying down an AI governance framework.
This article covers some of the themes covered in Manuel’s presentation at our Deep Learning Conference, be sure to watch the video to hear the full story!
Where are the success stories in banking?
It seems some institutions are trying to leverage AI technology. And maybe we’re wondering how come we haven’t yet seen any successful use of artificial intelligence. We have seen bits and pieces here and there, but we are yet to see that big success story in finance, like the ones we have seen in computer vision and healthcare in other places.
Banking vs Digital Native Disruptors
According to Manuel, “there are a lot of breakthroughs from a scientific standpoint. But the institutions themselves they only do not have the hardware or software or infrastructure but also the mindset that it takes to succeed commercially with machine learning..”
Rather banks are traditional, conservative institutions. Thus, they’re at a crossroad about where they have to leverage technology and artificial intelligence to catch up.
Manuel notes “It’s both easy and hard to create value with artificial intelligence. It’s easy in the sense that technology is out there. But it’s harder in the sense that it takes more than just hiring talent to get transformation going.”
In finance, there’s the myth that people tend to think that AI is just yet just another complex piece of statistics or analytics or automation. However, AI offers a step-change improvement in capability. In the way processes and services, are provided and how data is collected as well.
The perfect storm for disruption in banking?
Manuel notes that artificial intelligence and blockchain is evolving. “We also have clients and customers accustomed to a particular type of experience when dealing with the system due to their experience with Amazon, Google, and Facebook et al, and their skill at providing personalised services. There’s also a plethora of financial startups that are beginning to leverage technology and provide alternative ways of banking.
Manuel notes that while Canada is yet to lose the banking material market share to fintech entrants, other entries such as the UK have felt a tangible impact due to the increased competition from open banking regulations.
“In the UK, you have finechs now capturing 12% of the new clients or new revenue, whereas in Canada, it’s only 2%.”
What are the opportunities for AI in banking?
Marketing: a very personalised offering. “It provides a better understanding of our clients and their interactions with our digital platform. Thus, there’s a lot of data and metadata that is collected in that analysis to understand the customer’s profiles, but also their behaviour. This allows the bank to better target marketing campaigns.
Communication: great understanding of the client, call centers, chatbots, automated Q&A across all client interactions. Manuel notes this has been particularly important during COVID19 where clients are now needed support and access to emergency funds and getter access to customer support.
Decision and predictions: How can we use state of the art machine learning deep learning to make better predictions and understand fraud, risk, financial markets? How can I get better predictions?
“That’s one of the places where the mindset change is yet to occur. Because usually when you go into a team that does advanced analytics and they’ve been playing around with what they call ‘the AI’, basically what they’re just using a straightforward Deep Learning Network. And they use it instead of another type of model. What they do not see is that actually, deep learning is about knowledge representation, stacking different types of architectures, in a tailor-made way, such that you produce new features to represent a profile in a way that you can better predict and so there’s a lot of opportunities here.
Operations: the paperwork in the back office is significant. So natural language processing helps with documentation and to streamline paperwork in a more efficient way.
So what are the challenges?
According to Manuel is can be difficult to instigate change in banking, as it requires financial institutions to move away from their legacy infrastructure and “move towards some sort of a data exchange where data points can be exposed so that any team working with the data can just pull up whatever they need, they can train the model, the model outputs a new variable or insight that’s just integrated into the data exchange so that all the team can have that information.”
There are also organisational challenges. AI requires a change in the organizational mindset and an understanding of the organisational impact of change.
The social impact: “When you deploy something new and you target artificial intelligence, you have to have the governance framework that backs you up and can ensure that doing it the right way. You also have to reassure clients that you are respecting data privacy and acting ethically responsible.
AI facilitates new ways of interacting with clients and creating a feedback loop. As Manuel notes, “The more we interact with a client, the more data we’re collecting, and the more our models can represent that knowledge and use it somewhere else.”
New technologies often often often face reluctance on the part of the decision-makers and that’s even more true in banking.
There were lots of questions from participants, you can watch the rest of the video posted above to view the discussion.
There were lots of questions from the conference participants, you can watch the rest of the video posted above to view the discussion.