Inside Ocrolus: AI features making a difference in mortgage

HousingWire

What does AI in mortgage really look like in action? Rebecca Seward joins Digeo Sanchez to explain how AI isn’t there to replace underwritters or loan officers — it’s here to amplify them. She shares insights on how AI can streamline processes, reduce repetitive tasks and helo mortgage professionals focus on higher-value work.   Rebecca also explores how AI can drive long-terrm industry stability by distrupting the traditional boom-and-bust workforce cycle and supporting a more sustainable staffing model. This conversation breaks down practical AI strategy, showing how the right tools can surgace better insights and help teams work smarter — not harder.

Transcript

00:00We are live from Las Vegas.

00:08I'm Diego Sanchez, President of HousingWire,

00:11and I'm joined today by Rebecca Seward of Ocrelis.

00:15Rebecca, it's so great to talk with you today.

00:17Likewise. Thank you.

00:18So before we jump in,

00:20could you briefly introduce yourself and Ocrelis?

00:23Of course. My name is Rebecca Seward.

00:25I am Head of Mortgage Products at Ocrelis.

00:28We are an AI-powered data analytics company,

00:32is what we like to say.

00:33Of course, we do document automation.

00:36We leverage AI, ML, OCR technology,

00:40plus human-in-the-loop,

00:42to extract data and generate AI-empowered data analytics

00:48to drive faster, better decisions for lenders

00:51across multiple industries, including small business,

00:56but also, of course, mortgage.

00:58So we are here to embed ourselves

01:01into the digital lending ecosystems

01:03to help make faster, better decisions.

01:06Yeah, that's great.

01:07And AI is definitely a buzzword

01:09that you're hearing in the hallways here.

01:12How do you think about AI in practice

01:16as opposed to it just being a buzzword?

01:19Yeah, of course.

01:20So for us, we believe that AI is essentially

01:24an augmentation tool that enables lenders

01:27to make faster, better decisions.

01:29We do not see ourselves as a replacement

01:31for the human decision-makers.

01:33We believe that we have a part to play

01:36to enable them to be more efficient.

01:39So when we see the cyclical and seasonal nature

01:45of this market, of course,

01:46I've been working in this for 12 years,

01:48the ebbs and flows,

01:49we believe that you should utilize AI

01:53to enable your teams to scale more efficiently.

01:57And so I know there's this big buzzword around AI.

02:02It's not a random tool

02:06that just has all these different,

02:08I guess, pulls and levers, right?

02:12It's a human-empowered tool.

02:15The AI is only as good

02:17as the information that you give it, right?

02:22So Oculus, for example,

02:24is leveraging off-the-shelf tools

02:28like Google Textract and Amazon AWS,

02:32of course, or Amazon,

02:34but we leverage human validators

02:37to constantly empower our systems

02:40to get smarter and better over time.

02:42We feel the same way about the underwriters

02:45that are embedded in our workflow.

02:46So we actually read the information

02:49that they're giving us

02:50to enable us to get better over our analytics.

02:53So I think that humans have a big part to play

02:56in terms of making AI stronger and better

02:59and evolving it over time.

03:00Yeah, I mean, that's a big misperception, I think,

03:04that AI is going to replace humans

03:07as opposed to enhance humans.

03:09And it sounds like you're building product

03:12in that spirit.

03:13Exactly right.

03:14So I've been working in this industry for a long time.

03:17I understand the nuance of a mortgage underwrite.

03:20It is never as easy as just a simple person

03:24coming in with a wage earner income situation.

03:27There are now a lot of different ways

03:30of making income, right?

03:31There's Uber, there's TikTok, for example.

03:35And so we understand that there's always a part to play

03:39for the human underwriter or the human processor

03:41to inform the decision.

03:43So we are just here to make their job that much easier.

03:46We try to get them 90% of the way,

03:49present them with a decision-ready data point,

03:52and then enable them to make the necessary decisions

03:57to get to that credit decision,

03:59whether it was approval or denial or whatever else.

04:02So let's get really tangible.

04:05How do you help a lender either improve their processes,

04:11win more business, ultimately win market share?

04:14Of course.

04:15So we have a suite of products, essentially.

04:18I mean, we're an AI-first company,

04:20so we extract and classify documents,

04:241,700 different documents.

04:27So customers can, of course, leverage that

04:30for any data-driven decision point.

04:33However, our suite that is embedded within Compass

04:37does classification, essentially.

04:39So we are reading documents in from file manager

04:42or a listening folder within the e-folder.

04:44We classify those documents,

04:46and we place them into e-folder locations

04:49depending on the customer's unique taxonomy.

04:52So a human processor doesn't have to do that themselves, right?

04:55We standardize that across all of the documents

04:58that we receive.

04:59So every single time we see a bank statement,

05:02for example, we label that bank statement

05:04or we label an asset document, whatever else,

05:07depending on the customer's unique taxonomy.

05:09All the while, we've been extracting data

05:12off of those documents.

05:13And then we leverage that data to empower

05:17income calculations.

05:18So when a customer comes in,

05:20they see all the income calculations,

05:23whether it's wage earner, self-employed, rental,

05:26variable income, such as unemployment

05:30or retirement, for example.

05:33And then we also have a product that's in beta

05:36called Inspect, which does a data comparison

05:38of the Encompass 1003 to the document data

05:41that we process.

05:42So we can help expose things like undisclosed REO

05:46or a missing document for your initial needs list.

05:51So the history of mortgage workforce

05:57is one of boom and bust, right?

05:59We saw it in 21 and 22,

06:03big boom in hiring,

06:05and then 23 and 24.

06:07Unfortunately, big bust in hiring

06:10and a lot of people were let go.

06:11How do you think these tools

06:13that you're bringing to market

06:15help lenders not have such an accordion of hiring?

06:19Of course, it enables you to scale

06:21so much more efficiently, right?

06:23Every single time you see that refi boom

06:26or the surge in volume,

06:28a lot of lenders will handle that

06:31by hiring a ton of human collateral, essentially.

06:35So we believe that if you embed technology

06:38like Oculus or AI-empowered tools,

06:40it enables you to scale so much more efficiently.

06:43You essentially train your team

06:45to be 3x, 4x times more efficient

06:49to be able to process that many more loans,

06:51to be able to handle decisions

06:53that much more accurately and efficiently.

06:55And so you can essentially handle the surge in volume

06:58with the same amount of humans and human labor,

07:02but just enabling them to do it

07:04more efficiently through AI.

07:06You know, it seems to me that

07:08it's really important for lenders,

07:12especially the lenders,

07:13most lenders don't have huge tech teams.

07:16And so it seems like right now

07:18when we're still in the early days of AI,

07:19it's important to stay flexible with your AI strategy.

07:23How do you help your clients

07:26stay flexible with that AI strategy?

07:29Yeah, of course.

07:30So like I said, we involve our humans

07:32in our decision point.

07:33We're not here to present the decision

07:35and just allow you to accept it.

07:37We are here to present decision-ready data,

07:40but enable the customer to

07:43be flexible with that decision.

07:45So we, like I said,

07:49have a lot of humans in the loop,

07:53and we say that, a human in the loop.

07:55So every single time that we're seeing documents,

07:58we are learning from that.

08:00So we have a human flywheel, essentially,

08:03where we are seeing new documents come in

08:06and we see the human validation

08:08that's coming in from our verifiers,

08:10and it enables us to get that much better

08:13and smarter about our AI models.

08:16We feel the same about the customer involvement.

08:19So we actually read the different edits

08:21that they're making,

08:22the different ways that they're using

08:24our income calculation,

08:26and we get smarter about that over time.

08:29In terms of the human involvement

08:31and the customer involvement,

08:33we go out on site.

08:35We train our teams.

08:36We have AI-empowered underwriter courses

08:39that we teach these underwriters

08:42about how they should enable their teams with AI

08:45and, again, not replacing you,

08:49here to make you better, smarter.

08:51And so in terms of flexibility,

08:53I think we are just trying to teach them

08:56of all the different ways

08:57that they should be embedding this

08:59into their workflow

09:00in terms of training teams,

09:02making it more consistent

09:05across all of the processing,

09:07the underwriters, the loan officers,

09:09so that when you're bringing in new talent,

09:11especially, I feel like the mortgage industry

09:13is, I've been saying all week,

09:16quite incestuous.

09:17It's a lot of people that have been here

09:19for a long time.

09:20So if we want to attract new talent,

09:22like the Gen Zers and the millennials,

09:23for example, that are fresh out of college,

09:26it's using these tools

09:28to actually create consistent learning

09:30throughout all of your new talent.

09:32Absolutely.

09:33And if you look back at that boom

09:36that we saw in 21 and 22,

09:38there was a real shortage

09:40of underwriting talent

09:41during that time.

09:42So maybe this helps with that.

09:43Yes, absolutely.

09:45And I think that you will find

09:48with all this younger generation

09:51coming on board,

09:52they will be looking for companies

09:53that are more innovative

09:55and are open to leveraging technology

09:57like Oculus and other tech vendors.

10:00Because nowadays, I mean,

10:02imagine the kids in high school

10:04using ChatGBT, for example.

10:07They are going to want to leverage technology

10:09like ours to enable

10:12more efficient decisions.

10:14Well, what an interesting conversation.

10:16Love talking about AI.

10:18Thank you so much for joining me today.

10:19Of course, thank you for having me.

Category

Transcript

Recommended