- 3 hours ago
Blue Sage AI
Blue Sage AI embeds intelligent document analysis, workflow automation, underwriting support, predictive analytics, and conversational AI directly into Blue Sage’s Digital Lending and Digital Servicing platforms to help lenders improve efficiency, reduce manual work, and maintain traceability across AI-assisted workflows.
Blue Sage Solutions
Blue Sage AI brings intelligent document analysis, workflow automation, underwriting support, predictive analytics, and conversational AI directly into the Blue Sage Digital Lending and Digital Servicing platforms. Built for mortgage workflows, Blue Sage AI helps lenders reduce operational friction, improve efficiency, and maintain visibility, traceability, and control across AI-assisted processes.
Blue Sage AI embeds intelligent document analysis, workflow automation, underwriting support, predictive analytics, and conversational AI directly into Blue Sage’s Digital Lending and Digital Servicing platforms to help lenders improve efficiency, reduce manual work, and maintain traceability across AI-assisted workflows.
Blue Sage Solutions
Blue Sage AI brings intelligent document analysis, workflow automation, underwriting support, predictive analytics, and conversational AI directly into the Blue Sage Digital Lending and Digital Servicing platforms. Built for mortgage workflows, Blue Sage AI helps lenders reduce operational friction, improve efficiency, and maintain visibility, traceability, and control across AI-assisted processes.
Category
🤖
TechTranscript
00:05Welcome to HousingWire Demo Day on Demand. I'm Alison LaForgia, and HousingWire Demo Days
00:11spotlights some of the most innovative technology companies in mortgage, giving the HousingWire
00:16audience a front row seat to real product demos from teams building the technology moving our
00:21industry forward. In this session, we're featuring BlueSage AI. BlueSage AI embeds intelligent
00:28document analysis, workflow automation, underwriting support, predictive analytics,
00:34and conversational AI directly into BlueSage's digital lending and digital servicing platforms
00:40to help lenders improve efficiency, reduce manual work, and maintain traceability across AI-assisted
00:48workflows. In this BlueSage demo, Joey will be taking us through BlueSage AI. Joey, the floor is
00:54yours. Thanks, Alison. Hi, everybody. I'm Joey McDuffie of BlueSage Solutions, and I wanted
01:02to introduce a little bit about our product and some of the new initiatives we have going
01:06on, exciting new initiatives we have going on at BlueSage. For those that may not be familiar
01:12with BlueSage, we've been around since 2011, and our mantra was to build the most modern,
01:19dynamic platform in the industry. We've actually progressed a long way since 2011.
01:27And really, one of the things that makes us different is that we have what we call our
01:33digital lending platform, which is really the only multi-channel platform that has a combined
01:38POS, LOS, and even servicing now that's been written from the ground up in the last 10 years.
01:46Fully native. It's not a combination. It's not a Frankensteinian monster. We've written all of
01:51this from scratch, and we're excited to continue to add more and more customers to the platform
01:58that are realizing a lot of ROI. So from an AI perspective, since that's what we're focused on
02:07today, everyone in our industry seems like is focused on AI. BlueSage has actually been doing
02:15AI since 2022. We partnered with one of our current clients. We built some AI decisioning using ML
02:25models back in the day for HELOCs to do auto decisioning. And then since then, we've continued
02:31to add more and more features and functionality. In 2025, we added SageVision, which is what we call
02:39intelligent document analysis, which we'll go over in just a second. And then in 2026, we've added
02:45a few additional functionality called AI Studio and some AI agents to further streamline the process
02:53for customers. So let's jump right into see how SageVision works. As I mentioned before, we do
03:03actually have a POS. So in this particular case here, as the borrower, after I've submitted the
03:10application, you can see that this loan is in processing. And there is a number of conditions
03:16or a number of conditions that have been requested from this particular borrower. Instead of actually
03:23going in and associating a specific document to a condition, what we can do is we can basically just
03:29request a number of documents from the borrower. So in this case, I'm just going to call this all docs.
03:37And then we can add all of the files that are necessary from the borrower. So in this particular case,
03:44you can see that the borrower is uploading 1040s and pay stubs and driver's license and
03:49sales contracts and bank statements and you name it, right? All of those things can be uploaded in
03:56one fell swoop. So once we, once the borrower actually uploads those, then we actually put these
04:03through SageVision. So SageVision in a matter of minutes, what it'll do is it'll actually go in,
04:12analyze each one of these documents, both for what the document type is, as well as start doing
04:21some extractions that can be used further in the process, right? So not only do we just go in and
04:28say, is this a 1040? We can go in and actually check to see if this is Alice's 1040. Is
04:35it the right
04:35one from last year, the right year, the right employer, you name it? And that's where all of
04:42the AI capabilities come in. So in a matter of minutes, once these documents are processed,
04:50then we can actually go over and see what happened in the LOS. So I'm going to flip from the
04:54POS over to
04:55the LOS and go and see what happened behind the scenes. So in this particular case, I'm logging in
05:03as PADI processor. So I'm going to log in and find that particular loan. In this case, here's my loan
05:12here, 5543. All of those documents, I think there were 14 documents that were uploaded, all actually
05:19funneled into what we call our document manager. So in this case, you can see that while Alice,
05:28that simulated bar, were uploaded all of these, these were all indexed appropriately based on what
05:36was being uploaded at that particular point in time. And not only were all of these classified,
05:42you can see that we also provide an extraction panel that has various data insights that are
05:49gleaned from that each individual document, each individual page of the document.
05:54One of the key aspects of that is that we have this confidence factor, right? So as the
06:00intelligent document analysis engine called SageVision is processing each one of these,
06:06we can actually go in and see not only where did that information come from on that particular
06:12document, but is it enough to actually feed something else to automate the process even further?
06:19So we can do this with any number of documents, whether it's a bank statement, or like I said,
06:23a driver's license. And that information can then be used by agents and other aspects of the loan
06:31process. So in this particular case, for example, if we were to go look at the driver's license,
06:37we pulled off the expiration date, the date of issue, the name, et cetera.
06:42So before we go in and look at a couple other things, you can see that once the borrower uploads
06:48this information, we can pre-populate the loan application as well. So instead of in the old days
06:54with the processor having to go in and look at the image and then come over to the screen and
06:59say,
06:59okay, this is Alice's driver's license. It's not expired. Once it passed all of those things,
07:04as long as it was a valid driver's license, we actually take the engine and pre-populate this
07:09information. Same thing with the sales contract, right? So in the sales contract, we can actually
07:15pull off the realtor or any other information. So really any part of the loan application can be
07:20pre-populated and streamlined using this approach with SageVision. Now, in addition to SageVision,
07:27one other thing that we've added in the last few months is something called AI Studio.
07:32So AI Studio is basically another tool that is used for pre-underwrite. As I like to say,
07:41it's job enhancement, not job replacement. It's not an auto approval engine. It's really a more of a
07:47pre-underwriting feature designed to apply agency guidelines, portfolio guidelines more consistently
07:53without taking the judgment away from you, the underwriter. So I always like to say that it takes
08:01the scavenger hunt out of the equation, right? So there's two components to AI Studio. One of them
08:06allows us to go in and use loan data to create scenarios. We run that through the AI guidelines
08:13provided, and it will return based on the guidelines for that particular product, whether or not that
08:20loan is qualified or not qualified. This gives the ability for the underwriter to go in and really
08:27sandbox this particular loan to not affect the actual loan data, but to go in and say, well,
08:32what if we pay off this particular item? And you can see that the AI engine actually came back with
08:39a
08:39few recommendations for approval, and that's all going through AI. The other piece of the puzzle with
08:45AI Studio is that we have a number of conditions that were auto assigned to the loan. The AI engine
08:54actually goes through and figures out what is needed to satisfy those conditions, what documents
09:00are needed, and then once we actually do the analysis with a combination of SAGE vision and
09:06agents, you can see that it'll come back and say, well, our proposed status is to auto clear that
09:12particular condition. It auto associates the images with that particular condition and then gives us a
09:18finding, right? So in this particular case, you can see that all of the data points matched for the
09:24W-2, including the name, the social security number, et cetera. The underwriter can go in and review that
09:29there. But in the case of the bank statement, you can see that they, the AI model actually revealed
09:37multiple undisclosed recurring liabilities, including a payment. The direct deposits did not match the
09:43employer that we have on the loan. So again, that's a pinned. So the idea is that we can create
09:49any number of
09:50agents to analyze documents as well as data that we find from verification providers and other providers
09:56to go in and streamline the loan process. So the AI studio allows us to go in and see how
10:04each one of
10:04these agents were evaluated. So basically, almost like your sixth grade math teacher wanting you to
10:11show your work, we have that analysis based on some of the agency recommendations to go in and make sure
10:18you
10:18actually arrive at that conclusion. And then finally, the last piece of the puzzle, not only do we have
10:25all of this in our AI studio, but we've also moved this into loan conditions to make it very simple
10:32for
10:32the underwriter as they're going through and evaluating conditions. You can see that we update the status
10:39automatically if there's an issue with the content that they provided, as well as those findings that we
10:46had from the AI studio. So that makes it very easy for the underwriter to go in and understand exactly
10:55what was evaluated, what the findings were. And again, we can go through and do auto clearing of
11:03conditions and auto escalation of conditions. So in that particular case, we've gone through and
11:11looked at a little bit of Sage vision, we looked at AI studio. And the bigger piece of the puzzle
11:18with
11:18Blue Sage is that what we're finding with our all of our clients is that we have native AI built
11:26in. So
11:27we're continuing to provide additional innovation and value to our customers, they don't have to
11:33actually go out and get another vendor. And the biggest thing is that since we're the system of record,
11:38both on the POS, the LOS and the servicing side, these agents can operate on the data as this data
11:46is
11:46continuously being updated via more information from the borrower, third party services. And again,
11:54there's no data loss. Lastly, there's fast data ramp up to add agents. And, you know, don't wait,
12:02scan the barcode for more information.
12:03Joey, thank you so much for taking us through Blue Sage AI. I have a couple of questions for you.
12:12Can you explain how Sage Vision handles a real document scenario end to end, including confidence
12:19scoring, source references, and what happens when a human review is needed?
12:24So really, you know, what I always like to say is Sage Vision is your grandfather's
12:30automated document recognition and data extraction engine. It's amazing how much technology has come on
12:39in the last few years. So what Sage Vision gives us the ability to do is really analyze a particular
12:46document, extract data, virtually 100% of the data. And then, but where it may have some questions,
12:57we can actually apply that confidence score. And that confidence score, if it's all above a certain
13:03threshold that the lender can, that the lender can define, then we can say, we're going to do something
13:08with that. We're going to clear conditions. We're going to automate the process, progress of the loan
13:15to the next phase. We're going to actually go clear to close. If for some reason, the confidence score
13:20is such that it's below a certain threshold, then we can actually do auto escalation and auto tasking
13:26to a specific party on the loan so that there's further follow-up on that particular document.
13:33You had mentioned that Blue Sage AI is embedded directly in the platform. How does data flow
13:40through the workflow and how do auditability, traceability, and customer data protection work?
13:47I think that's really one of the benefits of having native AI within our platform versus actually adding
13:54a bolt-on or a plug-in from some other service. All of the data resides in each lender's environment.
14:02So nothing's actually being, you know, nothing's leaving their protected environment with all of
14:09the security aspects that every lender expects. From an auditability perspective, with all of our AI
14:17and agents, we basically show the work, right? So each one of these agents, we have the ability to do
14:26configuration. So all of the lenders' guidelines, et cetera, can be configured to reflect their policies.
14:37And then as these agents are processing that data, we have full traceability and auditability
14:44so that, you know, based on Freddie and other agency recommendations that you have a record of exactly
14:52how you arrived at that decision or that review without actually basically just using a calculator
15:00and saying two plus two is four, right? We know how we got there and what happened behind the scenes,
15:06what documents were used, what data was used to arrive at that decision, whether it's a pass
15:11or it's a fail or it's a review. And Joey, how do lenders get started with Blue Sage AI and
15:18what capabilities are actually available today versus ones that are planned roadmap enhancements?
15:25So that's a good question as well. So, you know, again, for any of our existing clients,
15:29it's very simple to turn on these agents. You know, some of these can be stood up in a matter
15:35of
15:35days. We've actually created these agents so that they are LOS agnostic. So we have about,
15:42I think we have about eight or 10 agents today. We're continuing to add others as we speak.
15:49And again, like, again, we can use data from documents. We can use data from third-party
15:54services like verification, BOI, BOE providers. And we continue to add agents through our portal,
16:02our agentic portal, you know, every week to further streamline the mortgage process and the,
16:10you know, the verification process to get, you know, to get your cost per loan down and to get
16:16clear to close faster. Joey, thank you so much for joining me today. To our audience,
16:22for more information about Blue Sage AI, click the link below.
Comments