Loraine Lawson:
Good day, I’m Loraine Lawson, associate editor for Bank Automation News. Recently, I spoke with Sean Hunter, the chief information officer for OakNorth. OakNorth is a commercial lending credit intelligence vendor, which also happens to operate a commercial bank of the same name in the UK. We talked about what automation trends he’s seen and how technology like AI and open data may play out in the commercial banking space.What are you seeing in commercial lending in terms of automation?

Sean Hunter
I think they, you know, there’s been a real acceleration of the sort of move to adopt more automation. In commercial lending, I think in your pre-crisis, there was already a sort of a trend, but it was moving kind of slowly. I think the crisis is sort of emphasized the need for banks to move faster, use more data, make better credit decisions at the origination point, but then also better monitor the portfolio to kind of figure out what’s going on. And I think those are all things that were true before the crisis, but the crisis is really like, increased the intensity of the of the sort of speed and so

Loraine Lawson
Are you seeing them use AI? in commercial lending at all?

Sean Hunter
Yeah, definitely. I mean, we use we use AI cells as part of our solution. And we are seeing adoption of AI, in in a few different areas, across commercial lending, I think, you know, obviously, there are challenges, and there are things that, you know, regulators are interested in, you know, explainability, and so on to make sure that creating outcomes or, you know, it can be seen to be fair and compliant with the various laws and regulations. But yeah, guys, I mean, I think it’s a part of a toolbox for making things more efficient, using data to make better decisions in June.

Loraine Lawson
So what areas when people automate commercial lending, where do they focus on is typically based don’t automate across the process? Right? So I’m wondering where do they focus their automation when they’re doing it?

Sean Hunter
Yeah, good question. I would say, there are two main thrusts that we see in the market. And then there’s what we do, which is a third one altogether. So I would say the first one, which whichever and sort of started with is automation of various parts of the lending workflow. So if you think about the process of lending, there are a lot of parties involved. There’s lots of documents going backward and forwards in the process historically has been very inefficient. And a lot of people focused in the first quarter on how can we automate that digitize that as much as possible. And I think that’s a valuable piece of effort to make, I think the second piece of automation that we’ve seen, is automated decision making. And we’ve seen some of that, especially in the lower end of commercial lending. So if you’re talking about this sort of funding circle Lending Club, hi, Walker, cabbage, those people, and that tends to be, you know, pretty small loans, especially ones which are quite commoditized. So things like, you know, asset financing, invoice discounting, where it’s actually relatively straightforward to make a decision or where the ticket is size is small enough that the bank is prepared to almost take a credit card type of approach. And that’s the only approach that really make sense because the ticket size is small enough that you can’t really afford to have humans involved.
Our approach is a little different, as I said, because we’re looking at loans which are large enough, where full automation would be very uncomfortable, even for relatively big bank. If you’re talking about loans of you know, a million dollars plus, you can’t really take a credit card type of decision, there isn’t enough data to fit such a model for loans that are similar enough because each loan is its own special snowflake in a way. And then. So that’s the one thing and the second thing, you can’t take a fully manual approach these loans like you might do for a white-glove service for very large corporate lender, because at that end of the spectrum, obviously you put our whole team you work on it for a few weeks, etc. But you can’t afford to do that for all $2 million loans.
So our approach is somewhat of a hybrid of these two things, what we use is automation to do various parts of the process, make it more efficient and so on, but keep the human in the loop. So, what we do is we automate the process of the credit analysis. So for each borrower, we produce a sub-sector-specific credit forecast, you know, we calculate Debt capacity, liquidity, capitalization, etc, all this really important stuff, build a financial model natural, fully automated, and it’ll be for a specific sub-sector very granular. So, you know, the model for a restaurant wouldn’t be anything like the model for a hotel. And in fact, the model for a business hotel won’t be the same as for a wellness bobrick type of hotel, you know, we, we go very granular. And then we also automate the process of peer analysis. So figuring out in your own portfolio, what companies are similar to a given bar, how are they performing? How does my borrower stack up against those companies. And then the final piece is sector. And so looking at the wider sector, and we automate the process of gathering the most important insights. So say, I want to lend to a wellness type of hotel, you know, what are the most important things driving that industry? And, and what is the trajectory of those things and getting all the data so we do all that. And then for an existing portfolio, we do what we call portfolio insights. So we use these scenarios that we built up to predict the future debt capacity, liquidity profitability of these businesses, and then essentially, calculate a vulnerability score started with COVID. But now it’s just in your, in general. So how is this business kind of fare over the next 612 months? And you know, which are the businesses that our bank should focus on? Because the problem with the SEC is they do need a certain amount of attention because they’re, they’re relatively large lines, but you can’t afford to, you’ve got enough of them as a bank, that you can’t really look at each of them equally, you need to know where to focus in so emser. That’s how we use automation.

Loraine Lawson
So outside of your product, do you see commercial banks doing any automations or use of AI that have surprised you or that you thought were innovative?

Sean Hunter
So I will say this, there’s some innovative and interesting work going on automating some smaller decisions, and using AI to to help with that. For example, I spoke to a Norwegian bank before they got their charter, and what they’re doing is they’re providing a service whereby they working with accountancy firms, and they’ll look at all of the invoices that a business has. And they’ll look at the cash flow of the business and say, Oh, we can predict the future cash flow for this business. It probably needs invoice financing. And we can offer finance. And then so when you as a business owner, go to your accountant. During that process, the accountant just gets a pop up saying, Hey, we think that they’re going to struggle for cash flow in three months time, you know, and we can give them a bit on their invoices right here. So it’s quite a, it’s quite a nice use of AI in the sense that the decision making is fully automated, but also, it’s providing the finance at the point of need, in a kind of universal is quite interesting way. So I think that’s quite interesting. I think there’s a few different, there’s a few different things. I think it’s a you know, it’s a field that there’s it’s ripe for exploitation, exploration, but are there?

Loraine Lawson
Yeah. Are there areas in which banks are resistance? Who automation? Do you think especially in the commercial lending space?

Sean Hunter
Yeah, sorry, I think. So I think there is resistance across the board in commercial lending, for some good reasons, right? So I think a good reason to be somewhat resistant, is, for example, explainability. So if you have an automated decision-making model, how do you know that it’s fat, for example. Now, obviously, banks have your federal and state requirements to be able to demonstrate fairness of money, especially if the borrower is in a protected category, and that kind of thing. But in any case, banks want to do it because they want to be fair. So the problem with AI sometimes is the models that it produces, while they’re very expressive in so they can capture some very interesting features or particular problem, it’s not always easy to understand the outputs of the model relative to the input. So Amazon, for example, at one point had an AI that was doing screening, initial screening of candidates who applied for jobs. And they famously scrapped that they brought it in because they thought, Oh, well, this will reduce bias in the initial screening process and help with diversity of our candidate pool. And then when they looked at the data, they actually found that the AI was somehow learning features, that was actually reinforcing bias in the application process. So I think those are good reasons to, you know, be thoughtful in take it slow.
I think our approach, the way we address those particular problems is we keep the human in the loop. So we don’t make a decision, we produce analysis that the human can use to make a decision. So what we want is, we want to help them to be the best credit off so underwriter that they can be, make really, really good decisions. And sometimes in commercial lending, you know, because of the things that I mentioned, like it’s slow, time consuming, etc, to do a detailed credit analysis, if you do it manually. You know, sometimes banks have optimized on easy decisions, rather than necessarily the best possible lend that they can do. And that tends to actually mean that certain bars are overlooked. So there’s lots of advantages for doing this type of thing. But yeah, that’s that’s the the objections that we’ve seen sometimes.

Loraine Lawson
So what do you think the future of commercial lending will look like?

Sean Hunter
So, I mean, we have a very strong opinion about that, actually. So our perspective, is we think the future of commercial lending is, firstly, in order to, like, really serve their customers beings need to take a forward-looking view. So that sort of previous approach of just looking at historical financials etc. Really, it doesn’t cut it. I mean, the COVID crisis really demonstrates that, because as soon as you get an event that’s not in your historical data, you’re really stuck. And your models don’t work and you’re incapable of acting.
So so we also think that banks need to get more granular, both in terms of sectors and in terms of taking a you know, sort of taking a loan by loan approach, rather than saying, oh, all my hospitality, I’m going to cut back on. There might be some hospitality, I’m picking on hospitality for some reason. But you know, all the loans in a particular bucket I’m going to kind of cut back on versus they might actually be opportunities in certain sort of snippets of a sector even if the sector as a whole has got problems.
We also think just in general, using data more, has got to help with making better decisions. Banks have always found it hard. They’ve got so much data But it’s very challenging to use. And we think that AI and automation can really help them with those things. I also think the future of commercial l is not just because the future of how digitization has worked in consumer lending has been like uberization – so can we make a really cool app that, you know, has got a great customer experience and it’s very easy to use. But I think for commercial lending, specifically, you need something more than that. Because, you know, business’s needs are very complicated. And, and so we think that technology can be used to actually give borrowers a more personal service, or more thoughtful service in a way that maybe they might have gotten in the past, you know, pre a, you know, I’m talking in the sort of 70s 60s, or whatever, when you had a local branch manager who understood your business, etc. You know, we think that like, with lots of data, with lots of really great tech, the, the relationship manager can really help the borrower. And so the bank, in the future state of commercial ending will be more of a partner, rather than just a lender.

Loraine Lawson
open banking is a huge topic, or at least in the retail spaces, if you think it’ll affect commercial banking, or do you see it out there?

Sean Hunter
It’s, it’s one of these things that everyone is there’s a lot of talking about open banking, and not so much at all implementation in commercial specifically, I think the place where it potentially has application, which might be quite interesting, is when banks once get banks get a bit better about using the data they already have, what open banking essentially unlocks is a lot of transaction data. So you know, you can also use the transaction data over customer to, for example, figure out, you know, what their cash flow is like, and all those kinds of things. And so if their cash flow currently has been impacted in a way that the financials aren’t good, showing open banking potentially can reveal that man, you can help them earlier. So we think that’s useful. But I think it’s also the area in which it’s potentially most transformative is the kind of use case like I mentioned, in Iceland, which is, you know, novel products and services, which are maybe not conventional banking as such, but by, by using banking data, plus other things you can offer the consumer, or the company in this case, you know, a better experience and better products.

Loraine Lawson:
Good day, I’m Loraine Lawson, associate editor for Bank Automation News. Recently, I spoke with Sean Hunter, the chief information officer for OakNorth. OakNorth is a commercial lending credit intelligence vendor, which also happens to operate a commercial bank of the same name in the UK. We talked about what automation trends he’s seen and how technology like AI and open data may play out in the commercial banking space.What are you seeing in commercial lending in terms of automation?

Sean Hunter
I think they, you know, there’s been a real acceleration of the sort of move to adopt more automation. In commercial lending, I think in your pre-crisis, there was already a sort of a trend, but it was moving kind of slowly. I think the crisis is sort of emphasized the need for banks to move faster, use more data, make better credit decisions at the origination point, but then also better monitor the portfolio to kind of figure out what’s going on. And I think those are all things that were true before the crisis, but the crisis is really like, increased the intensity of the of the sort of speed and so

Loraine Lawson
Are you seeing them use AI? in commercial lending at all?

Sean Hunter
Yeah, definitely. I mean, we use we use AI cells as part of our solution. And we are seeing adoption of AI, in in a few different areas, across commercial lending, I think, you know, obviously, there are challenges, and there are things that, you know, regulators are interested in, you know, explainability, and so on to make sure that creating outcomes or, you know, it can be seen to be fair and compliant with the various laws and regulations. But yeah, guys, I mean, I think it’s a part of a toolbox for making things more efficient, using data to make better decisions in June.

Loraine Lawson
So what areas when people automate commercial lending, where do they focus on is typically based don’t automate across the process? Right? So I’m wondering where do they focus their automation when they’re doing it?

Sean Hunter
Yeah, good question. I would say, there are two main thrusts that we see in the market. And then there’s what we do, which is a third one altogether. So I would say the first one, which whichever and sort of started with is automation of various parts of the lending workflow. So if you think about the process of lending, there are a lot of parties involved. There’s lots of documents going backward and forwards in the process historically has been very inefficient. And a lot of people focused in the first quarter on how can we automate that digitize that as much as possible. And I think that’s a valuable piece of effort to make, I think the second piece of automation that we’ve seen, is automated decision making. And we’ve seen some of that, especially in the lower end of commercial lending. So if you’re talking about this sort of funding circle Lending Club, hi, Walker, cabbage, those people, and that tends to be, you know, pretty small loans, especially ones which are quite commoditized. So things like, you know, asset financing, invoice discounting, where it’s actually relatively straightforward to make a decision or where the ticket is size is small enough that the bank is prepared to almost take a credit card type of approach. And that’s the only approach that really make sense because the ticket size is small enough that you can’t really afford to have humans involved.
Our approach is a little different, as I said, because we’re looking at loans which are large enough, where full automation would be very uncomfortable, even for relatively big bank. If you’re talking about loans of you know, a million dollars plus, you can’t really take a credit card type of decision, there isn’t enough data to fit such a model for loans that are similar enough because each loan is its own special snowflake in a way. And then. So that’s the one thing and the second thing, you can’t take a fully manual approach these loans like you might do for a white-glove service for very large corporate lender, because at that end of the spectrum, obviously you put our whole team you work on it for a few weeks, etc. But you can’t afford to do that for all $2 million loans.
So our approach is somewhat of a hybrid of these two things, what we use is automation to do various parts of the process, make it more efficient and so on, but keep the human in the loop. So, what we do is we automate the process of the credit analysis. So for each borrower, we produce a sub-sector-specific credit forecast, you know, we calculate Debt capacity, liquidity, capitalization, etc, all this really important stuff, build a financial model natural, fully automated, and it’ll be for a specific sub-sector very granular. So, you know, the model for a restaurant wouldn’t be anything like the model for a hotel. And in fact, the model for a business hotel won’t be the same as for a wellness bobrick type of hotel, you know, we, we go very granular. And then we also automate the process of peer analysis. So figuring out in your own portfolio, what companies are similar to a given bar, how are they performing? How does my borrower stack up against those companies. And then the final piece is sector. And so looking at the wider sector, and we automate the process of gathering the most important insights. So say, I want to lend to a wellness type of hotel, you know, what are the most important things driving that industry? And, and what is the trajectory of those things and getting all the data so we do all that. And then for an existing portfolio, we do what we call portfolio insights. So we use these scenarios that we built up to predict the future debt capacity, liquidity profitability of these businesses, and then essentially, calculate a vulnerability score started with COVID. But now it’s just in your, in general. So how is this business kind of fare over the next 612 months? And you know, which are the businesses that our bank should focus on? Because the problem with the SEC is they do need a certain amount of attention because they’re, they’re relatively large lines, but you can’t afford to, you’ve got enough of them as a bank, that you can’t really look at each of them equally, you need to know where to focus in so emser. That’s how we use automation.

Loraine Lawson
So outside of your product, do you see commercial banks doing any automations or use of AI that have surprised you or that you thought were innovative?

Sean Hunter
So I will say this, there’s some innovative and interesting work going on automating some smaller decisions, and using AI to to help with that. For example, I spoke to a Norwegian bank before they got their charter, and what they’re doing is they’re providing a service whereby they working with accountancy firms, and they’ll look at all of the invoices that a business has. And they’ll look at the cash flow of the business and say, Oh, we can predict the future cash flow for this business. It probably needs invoice financing. And we can offer finance. And then so when you as a business owner, go to your accountant. During that process, the accountant just gets a pop up saying, Hey, we think that they’re going to struggle for cash flow in three months time, you know, and we can give them a bit on their invoices right here. So it’s quite a, it’s quite a nice use of AI in the sense that the decision making is fully automated, but also, it’s providing the finance at the point of need, in a kind of universal is quite interesting way. So I think that’s quite interesting. I think there’s a few different, there’s a few different things. I think it’s a you know, it’s a field that there’s it’s ripe for exploitation, exploration, but are there?

Loraine Lawson
Yeah. Are there areas in which banks are resistance? Who automation? Do you think especially in the commercial lending space?

Sean Hunter
Yeah, sorry, I think. So I think there is resistance across the board in commercial lending, for some good reasons, right? So I think a good reason to be somewhat resistant, is, for example, explainability. So if you have an automated decision-making model, how do you know that it’s fat, for example. Now, obviously, banks have your federal and state requirements to be able to demonstrate fairness of money, especially if the borrower is in a protected category, and that kind of thing. But in any case, banks want to do it because they want to be fair. So the problem with AI sometimes is the models that it produces, while they’re very expressive in so they can capture some very interesting features or particular problem, it’s not always easy to understand the outputs of the model relative to the input. So Amazon, for example, at one point had an AI that was doing screening, initial screening of candidates who applied for jobs. And they famously scrapped that they brought it in because they thought, Oh, well, this will reduce bias in the initial screening process and help with diversity of our candidate pool. And then when they looked at the data, they actually found that the AI was somehow learning features, that was actually reinforcing bias in the application process. So I think those are good reasons to, you know, be thoughtful in take it slow.
I think our approach, the way we address those particular problems is we keep the human in the loop. So we don’t make a decision, we produce analysis that the human can use to make a decision. So what we want is, we want to help them to be the best credit off so underwriter that they can be, make really, really good decisions. And sometimes in commercial lending, you know, because of the things that I mentioned, like it’s slow, time consuming, etc, to do a detailed credit analysis, if you do it manually. You know, sometimes banks have optimized on easy decisions, rather than necessarily the best possible lend that they can do. And that tends to actually mean that certain bars are overlooked. So there’s lots of advantages for doing this type of thing. But yeah, that’s that’s the the objections that we’ve seen sometimes.

Loraine Lawson
So what do you think the future of commercial lending will look like?

Sean Hunter
So, I mean, we have a very strong opinion about that, actually. So our perspective, is we think the future of commercial lending is, firstly, in order to, like, really serve their customers beings need to take a forward-looking view. So that sort of previous approach of just looking at historical financials etc. Really, it doesn’t cut it. I mean, the COVID crisis really demonstrates that, because as soon as you get an event that’s not in your historical data, you’re really stuck. And your models don’t work and you’re incapable of acting.
So so we also think that banks need to get more granular, both in terms of sectors and in terms of taking a you know, sort of taking a loan by loan approach, rather than saying, oh, all my hospitality, I’m going to cut back on. There might be some hospitality, I’m picking on hospitality for some reason. But you know, all the loans in a particular bucket I’m going to kind of cut back on versus they might actually be opportunities in certain sort of snippets of a sector even if the sector as a whole has got problems.
We also think just in general, using data more, has got to help with making better decisions. Banks have always found it hard. They’ve got so much data But it’s very challenging to use. And we think that AI and automation can really help them with those things. I also think the future of commercial l is not just because the future of how digitization has worked in consumer lending has been like uberization – so can we make a really cool app that, you know, has got a great customer experience and it’s very easy to use. But I think for commercial lending, specifically, you need something more than that. Because, you know, business’s needs are very complicated. And, and so we think that technology can be used to actually give borrowers a more personal service, or more thoughtful service in a way that maybe they might have gotten in the past, you know, pre a, you know, I’m talking in the sort of 70s 60s, or whatever, when you had a local branch manager who understood your business, etc. You know, we think that like, with lots of data, with lots of really great tech, the, the relationship manager can really help the borrower. And so the bank, in the future state of commercial ending will be more of a partner, rather than just a lender.

Loraine Lawson
open banking is a huge topic, or at least in the retail spaces, if you think it’ll affect commercial banking, or do you see it out there?

Sean Hunter
It’s, it’s one of these things that everyone is there’s a lot of talking about open banking, and not so much at all implementation in commercial specifically, I think the place where it potentially has application, which might be quite interesting, is when banks once get banks get a bit better about using the data they already have, what open banking essentially unlocks is a lot of transaction data. So you know, you can also use the transaction data over customer to, for example, figure out, you know, what their cash flow is like, and all those kinds of things. And so if their cash flow currently has been impacted in a way that the financials aren’t good, showing open banking potentially can reveal that man, you can help them earlier. So we think that’s useful. But I think it’s also the area in which it’s potentially most transformative is the kind of use case like I mentioned, in Iceland, which is, you know, novel products and services, which are maybe not conventional banking as such, but by, by using banking data, plus other things you can offer the consumer, or the company in this case, you know, a better experience and better products.