Whitney McDonald 13:33:55
Hello and welcome to the buzz of bank automation news podcast. My name is Whitney McDonald and I’m the editor of bank automation News. Today is May 14 2024. Joining me as co founder and chief revenue officer at arteria AI Abrar. Huck, he is here to discuss the importance of accessing and organizing unstructured data within financial institutions. And what role AI can play in helping a Pfizer accomplish that. Prior to co founding arteria AI in 2020, Abrar spent more than six years at Deloitte, Canada as a senior manager and operations lead in AI and analytics. Thank you for being here.

Abrar Huq 13:34:28
Absolutely. Thank you for thank you for having me. So my name is Abrar. Hawk. I am one of the co founders of Archer AI and the chief revenue officer. And what we do at arteria, is we help financial institutions really harnessed their structured data in particular out of their documentation processes, there are so many points in a customer relationship or a bar relationship, where an FYI is dealing with documentation, that documentation carries such important structure data that could be used for so many downstream purposes, and so many operational processes. At an FYI, we structure that data at source we we use AI to harvest that data that’s typically unstructured in documentation and turn it into structured data that can then turn into to value add value of creative and usable data for any number of purposes that I’m sure we’ll get into in the course of this conversation.

Whitney McDonald 13:35:35
Maybe you could take that a step further, real quick and define structured data versus unstructured data.

Abrar Huq 13:35:42
Yeah, for sure. So a structured data is data that that’s, that’s organized into individual data elements, if you will, and fits into sort of a database structure. And it’s really the operational lifeblood of how how FIA is operate in modern time. The challenge is that in the world of data, only, only 5% of data is is typically structured at the outset. And a lot of it is unstructured, meaning that it is it’s not digitized in any form. It sits in, in our case, it sits in documentation, it sits enough, a PDF or a Word document that’s not organized in a way that a database can consume, or that any sort of tooling Can, can can leverage it in order to, to answer another question or a subsequent question or drive another task or another behavior. And so that’s a lot of data that’s unstructured. 95% of data is unstructured. That’s a lot of lost opportunity for RFIs. They’re basically making decisions and driving insights off of a fraction of the information and data that’s available to them. And that’s when you think about it, it’s pretty crazy, then that, that effectively 19 out of 20 pieces of information that are potentially available to you are being ignored or lost. And it may not be that you’re ignoring it per se, but you just have no way of getting at it.

Whitney McDonald 13:37:32
Yeah, when you put numbers to it and say 5% is structured on the onset, and the rest is that 95% You kind of have to go out and get it you have to find it, you have to organize it. That’s that’s a big number. And that should be meaningful to our audience as well. We’ll get into what role AI can play in helping tackle that 95%. But before we do, let’s take a step back. Let’s start bigger picture here. Let’s just talk through the importance of understanding your data within a financial institution. What role does data really play within financial institutions? And I’ll just have you kind of start there and tell us about the importance of understanding your data. Yeah, sure.

Abrar Huq 13:38:19
It’s, it’s one of the reasons why we call our company arteria. We think of our ourselves, and documents as being the arteries that carry the lifeblood of an organization. And to us, it is that lifeblood, particularly for for FIS, it’s fundamental to, to everything from Are you going to offer a loan to a particular borrower or more customer to understand and how you interact with a particular or principle of customer in multiple areas, if there’s multiple products that are offering or multiple services that you’re providing to that customer, getting a holistic view, and then also around information sharing, and really making data accessible to, across an organization, particularly in a world where FIS are becoming more and more complex, they’re becoming more and more national and global. There’s, there’s an importance of ensuring that there’s consistency and standardization and processes, firstly, but also information sharing and having a common view of of a particular portfolio or of a particular customer. If you don’t have the data around that, you’re not going to be able to serve your customers appropriately or the best way that you can. And so it’s really fundamental to what f5 is do in a lot of different facets.

Whitney McDonald 13:40:02
So let’s break into the how of this. So maybe you could tell us a little bit more about how AI can be leveraged to tap into this data into this unstructured data.

Abrar Huq 13:40:15
Yeah, so there’s been a lot of advancements in AI, obviously, recently. And that’s one of the things that’s really unlocking the unstructured data to be able to create usable, structured data that can go that can go downstream and be used for other purposes. In particular, the forms of unstructured data that, that we’ve typically dealt with and in financial services, and banking are across a number of different mediums, they’re across a number of different levels and layers of complexity. And in particular, what’s exciting about about AI, the advancements in AI is how realistic and accessible it is to start to use AI and different techniques in AI to to organize that data and to parse unstructured unstructured data and turn it into structured data. And what that allows you to do is to make certain use cases that were that were future state or ambitious and in the past, to become areas where F eyes are driving and deriving real tangible value today. And so, as a good example, one of the big advancements in artificial intelligence in the last 18 months is that’s most exciting for us that are tackling this unstructured data problem is around multimodal models everyone’s been talking about, about generative AI and generally pretty obvious, obviously, over the last 18 months, but in parallel, there’s been a significant amount of work that’s been done to increase the the art of the possible with multimodal models, and that’s basically that think of it as, as a model that uses multiple senses of the human being. So, in the past, if you were to look at or analyze text and turn that unstructured data into structured data, you would use it would be very natural language focused and based off textual analysis. Now with multimodal, you sort of layer that in, combine it with, with, with visual analysis, spatial awareness of where particular sets of unstructured data might sit within a file, and combine all of that to into one model. And that has just opened up a new honestly a Pandora’s box of different possibilities that would have been very challenging, both from how achievable they are but also from a cost perspective of how do you actually run the AI or run the processes to to achieve those outcomes and make those outcomes more accessible and more fruitful at a much more accessible cost to which is a big part of the equation obviously, for for f5. In an environment like this where cost pressures are are very real.

Whitney McDonald 13:43:51
So if you’re a financial institution shinned that wants to leverage AI in this capacity? What is that take? What do you need to do? What do you need to have in place? If you wanted to tap into an arteria AI in order to better understand your data go through some of those documents or PDFs and pull data?

Abrar Huq 13:44:12
Yeah, it’s a it’s a great question. I think it’s, a lot of it is almost like a philosophical mindset. More than anything else. It’s. And as, as AI becomes a little bit more mainstream, and literacy and competency around AI amongst the general public and amongst bankers becomes more, more commonplace. It’s, it’s become easier for banks to realize value, because they are embracing a philosophical mindset of trying new things, taking bigger swings on media or problem statements. And that helps helps the dfis because they’re, because they’re taking bigger swings, they’re realizing more value than they would if they were just chasing after sort of low hanging fruit that had incremental value, but maybe wasn’t driving a significant transformation or change. Part of that is also in a mindset of understanding the limitations around what’s out there, whether that be AI or automation or combination thereof. It’s and there are still high expectations of AI and automation. And, and I would argue, much higher standards than anyone would ever put to a human being around that. And you, you can’t let perfection get in the way of progress on these things. And so the more that faiz are tolerant around the results that come from introducing an AI enabled process, or a new form of automation, the better off they’ll be now that comes with like adding, mitigate mitigates around the risks of that, whether that’s exceptions, management processes, humans in the loop, but there’s, there’s a whole host of thinking that needs to be done by the FBI around how you sort of put the belt and suspenders around the process and mitigate the risks that come with it. Which also can involve sort of thinking about setting up a slightly different operating model or a different labor model than that had previously been done for that particular task or process. But being comfortable around that and thinking through what those mitigations are, helps to, to, first of all, just do risk the the potential risks around around the introduction of AI or automation in a particular process, but also allows the dfis to move faster and just be able to withstand the risk was they mitigated and appropriately?

Whitney McDonald 13:47:23
Yeah, I think over the past couple of years, you’ve seen the the mindset shift around approaching Gen AI, in particular, you kind of moved away from we’re not using it, it’s too risky to okay, maybe we can use it this way safely. So you’re kind of seeing the mindset shift. Being able to recognize, yes, there’s opportunity, but there’s also risks. So how do we kind of meet in the middle? Maybe we can move in here to an example. I know at the top of the call, and you’ve kind of alluded to it a couple of times. You can you can tap AI or technology to understand or pull data from Word documents or from PDFs, maybe you could walk us through an example of what sort of meaningful data might be held in those in those documents.

Abrar Huq 13:48:13
Sure. A lot of it is actually around straight through processing for us all use the martir AI examples in. In this case, when you look at let’s let’s talk about commercial lending as an example. There are a number of different sub processes within loan origination as well as loan servicing that touch documentation that leverage and use and should use the same set of of structured data throughout. But in fact, that doesn’t happen because you are dealing with the documentation process, a sub document process here that goes deep with unstructured data that doesn’t tie into the next process, which doesn’t tie into the next documentation process. And so there’s a lot of unstructured data that results in manual keying and rekeying. So what you end up doing and how banks and FIS or lenders are using a product like arteria is, so at the outset, just firstly, around financial due diligence. So as a borrower walks into a branch or is, is shopping alone, that they’re looking for, to multiple institutions, they might be providing financial statements and all sorts of data. That is not actually structured data, they may not be providing Excel spreadsheets that you can then plug into a decision and process or adjudication system. But in PDF, financial statements that are basically locked in that PDF, until in the old days, somebody goes in, opens the PDF, reads the document, analyze that maybe accuser and a few things into their loan origination system. With AI, you can start to parse those documents that unstructured data into structured data that goes directly into the origination system, as well as feeds into all of the parts around the origination of the loan, whether it’s decisioning, spreading adjudication, all of the above. From there another, the net sort of documentation process that comes in is around the credit memo, where the very same data that’s coming from the financial statements, as well as some other data that that would, arguably already set within the loan origination system is feeding into a new document that’s getting circulated and collaborate on internally around the the approvals around the deal, and the covenants and non financial covenants that are that need to be put in place. And as that document is being worked on, credit amount has been worked on, things are changing within that document, as circumstances change as approvals get received, the data within it is changing. But if it’s in a Word document, or in that unstructured form, you never know how it’s changing until, in the old days, somebody goes and looks at it, again, can use it in all over then. But if you use a product like arteria, that’s, that’s turning that into structured data through the product, as people are working on it, you don’t have that sort of tuning and repeating your digitizing as you go along. And so those are just two examples in length. And that also then feeds into the commitment letters and the credit agreements, there’s a number of different sub processes, all of which are leveraging the same structured, that should be leveraging the same structured data, but they’re not because the processes are so deep in unstructured documents.

Whitney McDonald 13:52:09
That’s super helpful. Thanks for walking through a couple of examples of what you could really be pulling in. I think that when you when you think about the manual labor that it takes to go through and pull out data or even know where data is being able to look to technology to automate some of that process, or now a majority of that process is huge. I mean, you can’t listen to any earnings call or any any bank executive speak without hearing. Okay, we need efficiencies, we need time saving, we need to get rid of some manual efforts. So that kind of fits right into that. Um, last question, here is just one thing that the financial institutions that are listening here could could do in the short term that would help them tap into their data in a meaningful way. What would one takeaway be that they could do in the short term? Yeah,

Abrar Huq 13:53:00
I think one of the things that that would really be helpful for four or five years is to really be critical and to think deeply about the potential avenues of value that structured data can create. And I think the the burden that we’re faced with 95% of data being unstructured is that, that 95% has always been unstructured. And so we’ve never had access to that data and a lot of cases. And so we don’t know how to wrap our heads around what that actually means. But there’s all sorts of of, of ways in which the potential structuring of data within an unstructured data meant, could be helpful or beneficial for a downstream process. And it may not be directly adjacent to the documentation process, it might be two or three steps down in the workflow or in a, in an operational process, that that value is realized. And thinking through that. And understanding that is, is critical, because it also dictates how you approach the tackling the AI or automation process in the initial instance, where you’re dealing with it. So understanding and having a view of where you’re trying to go, will drive what you’re actually doing today. So really thinking critically and, and thinking about it in a deeper and better way than beyond just the operational efficiency of the process that’s in hand is one of the ways in which I have seen fit is really a drive exponential value out of this type of thing. There’s, in a lot of use cases and a lot of areas, there’s clear value in the immediate operational process that’s directly in front of you. But the real opportunity is and what does this mean? As a first or second or third order after this particular process that’s directly in front of me.

Whitney McDonald 13:55:26
You’ve been listening to the buzz, a bank automation news podcast, please follow us on LinkedIn. And as a reminder, you can rate this podcast on your platform of choice. Thank you for your time and be sure to visit us at Bank automation news.com For more automation news

Transcribed by https://otter.ai