Iron Mountain

Transcription:
Transcripts are generated using a combination of speech recognition software and human transcribers, and may contain errors. Please check the corresponding audio for the authoritative record.

Mario Palarca (00:08):
Good afternoon everybody. My name is Mario. I'm with Iron Mountain. Now you might be asking yourself, what is a 75-year-old company doing at a digital mortgage conference? I brought up some props just so everybody kind of reminds us of what we're known for.

(00:24):
Who sees one of these? Show of hands, sounds good. But in order to stay competitive in the market, you've got to evolve. Hopefully over time we get rid of all that and work together to build a more digital future.

(00:41):
So the problem with mortgages still remains the same, right? As much as we still push e-mortgages, and we're part of that journey as well, there's still a lot of documents in the mortgage industry. We know firsthand about 45% of all US mortgages are stored in one of our facilities. We track it, we monitor it, we QC it, and we drive the quality that we talked a little bit about earlier. Oftentimes you've got data and document discrepancies before you sell a mortgage, before you transfer a mortgage, or before you securitize a mortgage. There're all these checks and balances that have to happen in order to make sure that mortgage is enforceable, complete, and accurate. With that being said, Iron Mountain is here to help you with those painful processes. So, a little bit of a change from a buyer persona perspective.

(01:33):
We're not going after that home buyer. We're really going after that back-office operation-centric mortgage operations team that has to enforce and operate all these processes that you guys have to have. So we'll talk a little bit about Iron Mountain Insight, which is our flagship digital product at Iron Mountain. Regarding our pedigree, it's that document-centric perspective. Going in line with that, we built a best-in-class enterprise content management platform that competes at the convergence of content management and content repository; but also, as we embrace the advent of AI, we're also in there with you to automate and transform all the processes that you have to perform to enforce all those documents. Right off the bat, you'll see you have a customizable dashboard.

(02:30):
It provides the operator some best-in-class, quick informational settings, such as the number of exceptions that they've got to work, the number of ingested documents, the new loans that have come in, and so on and so forth. These dashboards are very much customizable. Depending on your role within back-office operations, you could adjust based on that perspective. When you take a look at our asset explorer, this is where your documents truly live. You'll be able to search them, you'll be able to chat with them, and you'll also be able to upload new assets as you clear out some of these exceptions. Right off the bat, you'll see that we have two different types of assets. One's called a case and one's called a document. A case to us is that workflow that you guys have to manage day in and day out. If you actually have an asset or a document that you've uploaded, you could upload that then and there. I'll just pick, let's say, a promissory note. You could double-click it just to see it. All the extracted metadata is on the side, so it's very easy to navigate. So this promissory note, for example, for Chelsea—nobody keyed that in. That's our system classifying the document and unlocking that document right off the bat. You can add some tags. So let's say like, "Hey Mario, take a look at this," if something's not right, and you could tag somebody. That kind of kicks off the workflow as well from the document.

(03:51):
You could also chat it. You could do some Gen AI here, summarize this document, or you could also interrogate it and have some questions that'll pop up. Hopefully this works—live demo as always. But yeah, right off the bat it's for loan number ending in 0959 for $141,000 and so on and so forth. Okay, let me give you guys a little bit of vision into the magic that's happening behind the scenes. This is our designer. This is kind of where you build some of the rule sets that we talked a little bit about earlier when you're trying to drive that completeness, accuracy, and enforceability. Oftentimes you'll get a loan package and you guys know it's 150 pages, 20 documents, and sometimes the documents don't necessarily match each other.

(04:45):
Sometimes they get inner-filed, sometimes they get lost. You could build some rules within the workflow that say, "Hey, make sure that the promissory note is always there. Make sure you're comparing against the LOS data or the delivery data to make sure that, let's say, the assignment and mortgage or the MIN is accurate or 18 digits." So right there and there you could build your rules so that the workflow is able to validate all these checks systematically. Additionally, you can also take a look at it from a classification and extraction perspective. This is how we take all the documents and we build these models for you. I'll just pick the promissory note just to stay in line with the debt concept here. As you could see right off the bat, you can have classification, metadata extraction, table extraction, transformation, and doc splitting from one screen in and of itself.

(05:38):
So that also reduces the doc prep that your operators have to perform. You can just digitize a whole mortgage and then the system can actually understand what document it is looking at, split it accordingly, extract the key pieces of information, and do all the validation down the line. Let's see here. That's how you kind of train it. That's our IDP. We talked a little bit about cases. Let me jump back. Let's say something didn't look right, so maybe it was missing a MIN, or maybe the data wasn't accurate between the document and the data. You jump into your task section and right off the bat you'll be able to see things that you have to work on. This is what our cases look like. This is a summary of a mortgage for $309,000.

(06:23):
There were four documents that needed to go with it, and there're some exceptions that that operator has to clear. That operator can upload documents, find related documents, or request documents directly to the borrower for remediation right off the bat. Last but not least, you could also do some generative AI from this perspective. You could chat with it, you could say summarize, or you could ask it what's missing. It's oftentimes much more valuable if you're looking at different cases—maybe you're working on a certain portfolio and you're trying to figure out where to prioritize all your work. You could actually chat across multiple cases or multiple documents. This is very important because as AI continues to get better, and all we're trying to do is reduce the hallucination that's happening, what the AI is doing in our platform is actually just interacting with the data and documents that you have for your specific instance.

(07:16):
It could go out to the web if you choose it to, but also if you wanted to just point it to the data and documents that are inside of your instance, that's all it's going to review, search, and analyze. Last but not least, we still have a physical nature of things at Iron Mountain. We also have a single pane of glass, and I'll share that with you briefly. You could search for your assets up here and let's say you're like, "Oh, I need that. I actually need that physical document." You could type in, let's say, a borrower name, click on it, and it'll search for that document in the metadata fields and pull it. This is an image of a document that's in our facility; so that's our single pane of glass. Then last but not least, I'll also show you our agents.

(08:04):
I think we are all trying to embrace how agents can really transform our work. I know I'm running out of time, I'm just looking, but this is where you can interact with an agent. You could ask it different questions. You can have it do analysis. Out of interest of time, I preloaded this where the prompt said, "Hey, do all the borrowers listed on a note match the names on the security instrument?" The agent can look at all those documents, understand if there're any discrepancies, and then provide a recommendation on what to do. If you want an operator to work on it, you say, "Hey, send this to julian@basispoint.com." Is that it? Hopefully you get an email to do all that. And that's my demo.