How COVID-19 is Impacting the State of AI in Banking

On this panel, industry experts (listed above) discussed The State of AI in Banking and how COVID-19 is affecting it. We’ve included a short transcription of the panel, beginning at 4:50 of the webinar.

Watch the full webinar here.

Eric Maloney, Blue Prism: So I have a two part question: what have you been seeing as far as the adoption and uptake of AI and ML use cases and adoption within banking up until the outbreak of COVID-19? And the second part of the question is, do you see any corollary to changes in adoption with the spread of the virus? In other words, do you see new use cases and new priorities because of the COVID-19 effort? So I can just start randomly. I’m going to start actually with Wayne.

Wayne Shoumaker, Wells Fargo: We’ve had a steady increase in the adoption of the incorporation of AI and ML models into model governance since Wells Fargo initially recognized AI and ML as distinct sets of methodologies that should be incorporated into our standard model risk management policy.

There’s really been a steady development of functionality that pertains to such models, but more recently because of the COVID-19, the attention on the economy and diversion of resources that might otherwise be spent on managing models’ impact analysis have been conducted and special attention has been given to what kinds of impacts should be escalated and what kinds of impacts should be watched. One of the interesting developments that we’ve noticed is an occasional model has surfaced whose purpose is to explore the epidemiological impact of COVID-19 on the institution.

So as part of the second line of defense, one of our roles is to make sure that not only does the model service intended purpose, but that it is sourced authoritatively. So if an epidemiological model is intended to express and AI and ML is used to characterize what might otherwise be expressed by a professional epidemiologist or a team of epidemiologists.

We need to make sure that the authority backs the model and that it’s not just coming out of a back office. That’s one of the roles of the second line. We need to ensure that professional authority is retained and not substituted by advanced computer technology.

Eric Maloney, Blue Prism: So anybody else want to give their ideas of where AI and ML have been growing in banking as far as popularity and where that may be shifting or reassessed because of COVID?

Roderick Powell, Ameris Bank: Yeah. Well, I have been seeing, up until COVID-19, continuing uses in the areas of anti-money laundering and financial crime fraud detection. The machine learning algorithms really lend themselves to those types of use cases and especially all the classification type algorithms. I’ve also seen more use of AI-type technology in robotic process automation or, as some people call it, intelligent automation. And even the bank our size, we’re not a very large bank.

We’ve embraced automating processes and a lot of these so-called bots use natural language processing to identify documents and classify text, words, and so forth. So I don’t think COVID-19 will stop progress towards adopting these tools. If anything, I think it will increase it.

As banks face some challenges with the economy in general. Obviously using RPA you can hopefully reduce expenses cause you’re automating human tasks and you’re leveraging some of this cognitive technology. I also see it being used in the marketing areas, in scraping websites, I’m seeing what your customers are saying about you. And again, this is a more natural language processing focus. 

As far as direct from COVID-19, all the banks had to adopt Cecil and that’s very model dependent. Some people don’t consider a regression analysis as machine learning, but it really is a form of supervised machine learning. Most banks are projecting their loan losses and obviously, because of COVID-19, people had to adjust their models to take the pandemic into account when they project their own loan losses so that’s been a direct impact in the modeling world.

Eric Maloney, Blue Prism: Rajeev, what are you finding at BNY Mellon? Are you seeing some new attention or some new needs for AI and ML?

Rajeev Sambyal, BNY Mellon: In my experience, the focus was always there on AI and machine learning, especially the last couple of years, right? It has gone mainstream. I think we mentioned intelligent automation, right? A couple of years ago, a lot of stress was around RPA, intelligent automation and AI went mainstream. Then people started realizing that there is some tangible benefit so it’s no longer a research area. We can actually use it in an applied sense and do that. So I think COVID has not stopped or changed the direction on that. So it is still very much a front row citizen on this. 

In fact with COVID, there is more focus from a business perspective and from a technology perspective, where AI and machine learning can help, largely due to use cases in marketing or even finance. All those discussions about areas where potentially AI can come in and help have been fast tracked rather than sitting in the back on the shop floor. 

I think it’s across the board, too. Not just specifically in banking and finance, but across the board. After talking to different people, it’s becoming more and more apparent that people who were skeptical about the use of AI and how models could come in and augment what people have been doing. I think those discussions have come to the forefront now.

Eric Maloney, Blue Prism: Sri, you have a unique perspective on this panel, being an open source vendor and the sheer number of clients that you engage with. What’s your perspective on this topic of where COVID-19 is impacting AI and ML consideration and adoption.

Sri Ambati, H2O: Thanks for that. I think H2O still has the world’s top 10% of Kaggle grandmasters, data science grandmasters. We have been privileged to be able to help the community, the healthcare community and community systems. We’ve been building a lot of models with a lot of the governments as well, starting earliest with China and then Singapore, Korea, and most recently, of course, in all of US in the earliest days. We’ve been building these models and starting to look at when it will peak and say what the likelihood is of the impact of social distancing and how that could help. So we’ve been looking at payment systems in India and here to see how we could figure out the impact of it. 

We happen to have the data scientists and we have the AI platforms. We quickly found out that the predictions around unemployment were pretty stark. Small businesses because the lack of foot traffic are vanishing as well. Forbearance and default become a real problem across the country. People started leaving the cities globally, even after the post, freeing Wuhan and all the big cities in China, I had an exodus of people to more sparsely populated places. You’ll see that that changes demographics, that changes how the world is going to look at a phone call coming from the middle of what used to be the flyover States, how banks were going to deal with mortgages in these countries. 

You’re also seeing globally that we’re working with banks in Australia and banks iIn every part of the world where governments are trying to fund small businesses through the banks. And so which companies have growing concerns how to land? Are they really claiming what they’re claiming? If they had 10 employees, are they 10 or are they really five? All of that analysis is a decision to be made through data. 

A lot of public data sources, and again, the data is fuzzy in many places. Even the death count is not exactly accurate. And so a lot of places where testing has backed up or backlogged like California has had to sort of test backlog and now they’ve come forward. But trying to get any proxy for the extent of damage and how long the damage is going to be. Can we learn from the cannibal transfer models from the 1918 pandemic where you have the first bump, then a bigger bump and a smaller bump, or it’s going to be as simple as two bumps.

A lot of different scenarios are being played out at different parts of the banks. We are obviously, as some of the audience asks and Wayne mentioned, we are still making sure that we are applying fair lending practices. 

COVID-19 has impacted a different rung in the society. It has hurt predominantly 70% of African Americans have been impacted. So is that a proxy variable for race? There’s lots of interesting questions that explainability and AI have to play here so expanding this model before quickly jumping into using a different mathematical techniques, but also simple techniques.

Roderick’s points are well taken. The vast bulk of models are simple regressions. The great mathematicians are using simple methods, not very complex methods that are difficult to understand. So the universities are reacting very rapidly to this as are healthcare systems.

 I think health is the number one reason insurance claims go through. Bankruptcy happens in the U S because of poor health. So I think they’re all very tightly connected. Health insurance and finance vertical supply chains are being reinvented as we know. Some of the largest cleaning supplies companies are thinking of where to put factories even. We call it BC for before COVID and the world has been reformatted.

And a wonderful world of new opportunity. Painfully humbling by just a 0.16 microns. We have been humbled as a species, but I think we’re going to come out of this very strong.

Tags   •   Finance

Recent Posts

View All Posts