JJ Hornblass
Hi everyone, this is JJ Hornblass and welcome to the Buzz from Bank automation News where we cover what’s happening in automation technology for financial services. This is our weekly wrap for what’s going on in the industry on May seven 2021. Before we get into our discussion, I want to thank bank automation whose advertisers Narmie and MX for their support, so thank you to them. And so pleased to be joined by Lorraine Lawson and Jaspreet Kalra of the Bank Automation News editorial team.

Hello to both of you. In general technology news. The New York State Attorney General’s office said that fake comments including 8.5 million of them funded by large us ISP is accounted for 18, approximately 18 million of the 22 million net neutral comments received by the FCC in two sec. 2017. That’s more than 80% came from this. Were from were fake comments. California regulators have indicated I’ve told Neobank chime that they must stop using the words bank and banking in their ads and chime has agreed to do so. And this news broke yesterday. And finally, Nintendo reported record full year profits for 2000. And the year ending first quarter 2020. of $6.2 billion net was up 88%. On a year over year basis. They had net sales of 16 point 1 billion up 34% on a year over year basis. And they sold 28 point 2 million switch units in 2020. Up 37.1% year over year. I’m not sure whether that includes sales to just breed or remain in bank automation industry news. The question that we’re asking is can data management itself be automated? And we we broke the story recently, that highlighted the extent to which there are gaps in data and data management in financial services. So let’s start with some of the basics. What are these data gaps? I mean, what are we talking about when we’re talking about data gaps? And and then maybe we can start talking about how some entities are trying to alleviate this problem in various ways. Maybe Lorraine, you could kick us off in this discussion.

Loraine Lawson
Sure. So I didn’t get specifics on what the data gaps are. But they are looking at the company that was doing that was aerospike. Stewart termi is the glooby, global director of financial services industry solutions. And he said they use AI to fill in data gaps. And they are doing unstructured data. So that might be an address that’s incomplete, I’m guessing here, but an address that’s incomplete or any kind of demographic data that might be associated with unstructured data. unstructured data is just data that isn’t in the table. So think about word versus say Excel. So that would be one way that they’re using automation to fill in gaps. Part of what caused us to look into the story is there are a lot of vendors out there who’ve started to claim that they automate data. And that’s that’s just not that’s just not possible. So it has caused me to look into the story. And what I found is that yes, there are ways you can automate the data process, but there really isn’t such a thing as automating data fully. You do need data scientist, or at least solutions that are designed by data scientists to manage your data.

JJ Hornblass
Just be what’s your sense for what we’re talking about here? Like where what is the what is sort of the underlying problem that’s being addressed?

Jaspreet Kalra
I mean, from have a 30,000 feet view, I’d say that it might be easier. to automate data collection now than it would have been 10 years ago, because now we generate data much more faster. Like say something as simple as tapping your credit card on a Starbucks counter is a data point. And if your financial provider is so smoothly linked to the API is that they can immediately transmit it to someone. Yeah, technically, you don’t need a human to extract that data, but to fully automate it, or just to gather it without any human role at all. I think we’re still far off from it, if not, like, making it completely impossible. But one question I would certainly have there is that, what sort of AI do they use to fill in these capsules? It’s like a guessing AI is it’s like a probabilistic AI? Well, how how have they built that system?

Loraine Lawson
Well, there are no SQL solution. So I don’t know. That’s a good question. Just breed, I would guess that they’re using some natural language and machine learning aspects there. But that’s a good question.

JJ Hornblass
And I mean, I think that the other question is, like, the extent to which this is, I guess, an issue or that maybe even aerospike is trying to address it? I mean, do we have a, do you have a sense for you, either you have a sense for scope, or, you know, like the degree to which you have insufficient or unclean data through, you know, in within financial services broadly?

Jaspreet Kalra
If I mean, so unstructured data is a big issue, because they unstructured data also comes in through something as simple as a chatbot. A person had interaction with a chatbot. It’s a text log, it’s not about what they talked about how long that took. So I read I read recently, the Deutsche Bank has been using blue prisms RPA systems to just take that unstructured data, mined it and put it into tables that can be processed by mostly by humans, or by a artificial intelligence system that can draw insights from it. But unstructured data by itself sort of is separated from incomplete data. I think that’s what’s the interesting thing might be to look in here. Is it just unstructured? Or is it incomplete? And is the AI helping you complete those gaps or fill in those gaps?

Loraine Lawson
Well, if you look at it, banks also have a great deal of unstructured data just in their transactional systems that is also unstructured data. So for instance, filling in gaps, like, where did they shop, you could you can pinpoint from transactional data where someone bought something, but maybe you don’t have the complete address, I come back to that example. Or you don’t have full democratic information on that customer that you might want to fill in. Those would be the sort of things that they specifically mentioned, wayfair, which they call the data science company, that that also happens to sell home furniture, they use a lot of transactional data with anonymized demographic data to fill in to make recommendations to customers for for instance,

JJ Hornblass
I also wonder whether you’re talking about circumstances where the usage of the data might change is so in other words, if you’re saying jasprit, for example, you mentioned the you know, kind of pls pls data on on what if you want to manipulate if you want to utilize that data, let’s say for underwriting, you’re talking about a different, it’s different segmentation different, it has different implications, as as compared to kind of at the pls level. It is, is that also where you’ve got, you’ve, you’ve got the need for, you know, significant, either refinement, cleansing it, Lorraine, to use the word that, that you used in your piece.

Jaspreet Kalra
I mean, you know, so one thing that even I have noticed across conversations with AI developers, is that a lot of the work that data scientists or the people who are building the system do is just cleaning the data, making sure it is in the format that can be used by the algorithm, and also making sure that it doesn’t get wrongly influenced by extreme values, that could give you skewed results. So it’s a lot of constant checking, while at the same time making sure that what’s being fed in goes in a way that is that produces the result you want the machine to produce, or at least move in the direction of producing the result you want to produce. So processing for sure is going to be a big chunk of the buy when it comes to automating these processes, especially data processing, and how much automated that can be also comes with a wrist extension of whether you can’t be completely hands free because if you’re completely hands free the machine is making those decisions on its own and There is a need for checks and balances. And,

JJ Hornblass
Randy, did you get a sense for like what’s driving? I mean, I guess just breed is kind of explaining or indicating why there might be a need for this automation of data. But with it, was there any, were there any other factors that came into that seemed to be at play that is kind of leading to these initiatives? Well, I

Loraine Lawson
think you have to look at what the data is being used for. And we are trying to use more depth for AI. So we’re kind of talking about that kind of use case, more than say, just oh, I want to know how much my customers spent that maybe you could find out in a traditional system. So I think, you know, AI is a factor, and I think just capability to automate more. And also, you know, for smaller bakes, they don’t have the staff, to staff, a data management team, or at least not one with data scientists. So they’re interested in solutions that can offer them sort of that same functionality as bigger base are getting with their data scientists team. And sometimes it is possible to get certain use cases down where the data scientist team has already, this is what you see vendors really offering sort of data, their data science team has sort of worked on a model. And it will provide this one functionality for you, because they’ve already trained the model, she gave an example of like photos of where a vendor had offered a solution where they have trained the model to look for fingers in photos. So that’s like 30% of photos, it turns out. So you can use that you don’t have to dump in the data, you dumping your photos, and it will sort those out for you. You don’t have to train that model. So that’s the kind of thing we’ll see vendors offering is very specific use cases where they’ve already trained the model and applied AI to it, rather than sort of a situation where you get automated and tell it to do what you want it to do. And it would automatically do it.

JJ Hornblass
I would I would I would expect that with time. The degree that models are trained will have a cart relation to their price from vendors in the market. You know, that’s I think your you know, maybe and I’m not quite sure I I’m not quite sure how you measure that in order to price it. But it would it wouldn’t you know what you’re saying Lorraine would imply that the more trained is the model? Most likely, the higher the price should be. And I don’t know that we’ve seen that, where you’ve got, you know, sort of different varying price points, based on degree of training, or have we?

Loraine Lawson
I don’t know, that’s a good question. I don’t know the answer to that.

JJ Hornblass
Something we can look into next week, I’m sure. What else are we looking into next week.

Jaspreet Kalra
So for next week, or like this week, plus next week, we have something on what sort of AI hiring trends, the big banks have been following with respect to how many positions they have opened, what sort of verticals they want to use things in. And also, it’s not just the big banks, it’s also retailers like the Home Depot, even agriculture manufacturing company, john deere, is now looking to expand its AI operations and sales, marketing and business insights. So the field is growing competitive while the use cases are growing. So that’s a trend that we’re going to be monitoring. And the other story that I’m looking into is that there’s been a recent rise in ransomware attacks across organizations. And we’re looking to see if banks and credit unions have been at the receiving end of that as well.

JJ Hornblass
Great. Thank you to both of you for your time and insights. And thank you to all of you for for joining us for this episode of the buzz. We look forward to seeing you next time. Please follow us on Twitter and LinkedIn and visit us at Bank automation news.com we’ll see you next time. Thanks