AI Governance in Mortgage? Start with MISMO FRAME

Marvin Stone (00:00)
This is Stewart in the Studio from Stewart's Thought Leaders. Hey everyone, thanks for joining us for another episode of Stewart in the Studio. This one's actually on location in Reston, Virginia, here at the MISMO Fall Summit. And I'm with some folks from MISMO who are gonna talk about FRAME today. Brian, let's start the introductions.

Brian Vieaux (00:18)
Sure, Brian Vieaux, president of MISMO. Glad to be here.

Amy Moses (00:22)
Amy Moses, vice president of strategic growth with MISMO, and thank you so much for having us, Marvin.

Jonathan Kearns (00:27)
Jonathan Kearns, vice president of programs and operations at MISMO.

Marvin Stone (00:31)
Alright, great, thanks. So, you know, we had great sessions the last couple of days here in Reston and really unparalleled excitement, record attendance and everything. And one of the things that really hit home was the new FRAME offering you guys have put together really to make compliance a lot easier for lenders and vendors, large and small. But I think really what we're thinking is while this seems like the start of something new, this was really kicked off by Fannie and Freddie a while back with their lender letters and notices to the lending community about really governing their suppliers and AI in general. So, you know, that's where we are today. Really we're gonna talk a little bit about what the FRAME framework means and what that is all about. So, Brian, do you wanna kick it off a little bit?

Brian Vieaux (01:19)
Sure, I'll start by saying where FRAME came from. Last October at MBA Annual, the residential board of governors of MBA, RESBOG, prioritized for its calendar year a deliverable around an AI governance framework.

And it was really meant to help lenders of all sizes, member companies of MBA, be able to responsibly adopt AI and not run afoul of any state regs or federal regs or any pending guidance from the GSEs, which came following that. And so, you know, once RESBOG prioritized it, they looked to MISMO to execute on it. And the team at MISMO, led by Rick Hill, frankly, at the time, really poured a lot of expertise and energy around construction, a framework, a V1, version one of a framework, that starts to address the systemic risk that lenders have as it relates to their use of AI.

Marvin Stone (02:16)
Yeah, it's been such a journey kind of going back to the service provider oversight that CFPB introduced years ago that really kind of stirred up a lot of this. The sending of questionnaires to vendors all the time and you have to answer them in a thousand different formats and things. So, Jonathan, you lived that experience. What was that like?

Jonathan Kearns (02:35)
Yeah, that was one of the reasons why we really want to do this is because, you know, in eClose world, we would get assessments and questionnaires from every vendor and it was you know, fifteen to thirty different vendors all the time, every year. Same questions, answered differently, formatted differently. So, we really wanted to create a template that the industry could use not only to take the information and assess it from a lender perspective, but also provide the information in a consistent formula from a vendor perspective.

Marvin Stone (03:03)
Yeah, it's been such a burden. I mean a thousand different questions and a thousand different formats coming at you every which way. So that's, you know, from the vendor side and Amy, you've talked to a lot of lenders about this obviously since introduction and it was just introduced recently. I mean it's not been out, FRAME hasn't been out very long, but what's been the reaction from lenders when you've talked to them?

Amy Moses (03:22)
Lenders are enthused, Marvin. I had a lender tell me just last week that FRAME is a lifeline for them. Especially for the small to mid-size lenders who don't have the internal resources to set up a large AI governance program, FRAME is that for them. And so, it gives them that tool that they can start to work with within their organizations. And you know, we're surprised we're talking to a lot of large lenders as well who are incorporating it into their AI governance programs they've already established. And that wasn't something we've anticipated, but we've had very favorable reception from large lenders as well.

Marvin Stone (03:59)
Yeah, and it's kind of interesting. I think yesterday we heard on stage from Fannie and Freddie on this topic.

Brian Vieaux (04:06)
Yeah, we didn't expect to get the reaction we got from Fannie Mae and Freddie Mac on stage. We'd have been conversations with them, you know, offstage, so to speak, but as part of a panel where we asked each of the GSEs to kind of send their in-house expert around AI governance in terms of how they look at their own AI governance. So, they sent Joyce Walsh from Fannie Mae and Daniel Miller from Freddie Mac. Jay Kingsley from Cotality moderated the conversation and, at least five times 'cause I wrote these down, there was like some mic drop moments and I think Joyce at one point made the comment that FRAME is table stakes from a lender perspective. Now this is you know, we're careful not to say that this is an endorsement or a Fannie Mae or Freddie Mac approval, but I think the signal from the stage was that lenders that adopt a program like FRAME, and FRAME in particular, are gonna be in a much better position, not if but when, they're you know, when the regulators are coming in or in this case the GSEs are coming in to ask how they're managing their AI.

Marvin Stone (05:12)
Yeah, it's so proactive instead of being reactive and waiting. You know, AI's been out for a while, but we're starting to see it in the vendor community. And there were a lot of questions that were brought up on stage about, you know, well, how do you know that somebody, a vendor is testing their model? How do you know which model they're using? How do you know if that's your private data is training the model, all those different things. So, you know, I guess I would say if the GSEs are at least, you know, pleased to see this coming about it seems like a great way to consolidate around this. But, you know, every lender's different, right? And every lender has sort of their own secret sauce way of doing things. So, they may have AI in-house and they may have vendors who have AI. So, Jonathan, when you see that kind of mix of technologies, does it get down to the use case level or is it just a particular technology? How does that work?

Jonathan Kearns (06:02)
Yeah, that's a great point where there's going to be a mixture and the first thing you need to do is understand where AI is being utilized in your entire organization, right? Because it's being utilized today. And so, you've got to identify those. And that's what the inventory helps you do is identify those. And it could be an internal product, it could be an external product, it could be a vendor product. And so that's where you really want to start with it.

Marvin Stone (06:27)
So, we've talked about FRAME, the need for FRAME and kind of you know, what lenders are faced with and what the GSEs would like to see, state regulators I'm sure, everybody wants to see that, you know, we're using AI responsibly where the consumers involved. So, you know, what is FRAME and what isn't it, for instance? I mean, because if I'm a lender or a vendor, how do I understand what FRAME is, what that brings to me?

Jonathan Kearns (6:54)
Go ahead.

Amy Moses (06:55)
Yeah, I was just gonna say, well, you can add on to this, Jonathan, but FRAME is an acronym for Framework for Responsible AI in the Mortgage Ecosystem. And it is a member, a MISMO member exclusive product. So, we encourage organizations to become a part of MISMO in order to access and download FRAME. And you know, to Brian's point, we created this because of industry need, which is what really MISMO is all about, right? Listening to who is in the industry, what their needs are and then creating something that is a standard or in this case it's outside of a standard, right? It's not a standard, but we knew that this was a tool, a resource that the industry needed. And you can add on too Jonathan, what isn't FRAME?

Jonathan Kearns (07:39)
Yeah. So, like she said, it's a template. What it isn't, is it isn't the thing that you just put in and don't work with. Like you still have to put in the work. It's got a lot of information in it already, a lot of sample information and a lot of general guidance. But you have to work through it with your organization and ensure that your governance policies are adhered to this and following the rules because it's really ought to your point, it's at an individual organizational level.

They'll deal with things differently. We saw that in the FRAME workshop when we went through it. The same use case. There were four different groups and none of them came up with the right, the same risk tiers, you know?

Marvin Stone (08:19)
Yeah, that was fascinating during the workshop and that was such a great exercise because different people would see, in the same work group, would see things differently. There was one work group that got pretty heated about whether something was high risk, medium risk or low risk. And I'm not sure exactly who won that fight, but it was pretty interesting. So, FRAME has a lot of educational materials included in it. And it also has these templates that you can sort of, I don't want to say paint by the numbers because it's certainly not that easy, but you can fill out the templates and those are pretty complete. I mean there's really not much there. But if somebody didn't have something, they could add on to it in some way if there was something special.

Jonathan Kearns (08:57)
It’s a good baseline to start with. And it gives you a lot of information. And it gives you the ability, instead of starting from a blank canvas, to start with information on the discussions, to have the discussion with your governance committee, right? We all have—every lender has to do governance today. This is just another way. We have to understand how we got to the decisions that we got to. And now utilizing AI in those decisions and in our work process, we have to understand how that's being utilized and that's what this does.

Marvin Stone (09:27)
Yeah. That's great. So, Brian, you have experience across lending for many, many years, different organizations, large and small. You know, I think we all know that there are lenders out there that have a compliance person, and then there are lenders that have a compliance team and then a compliance division. So, it's across the board. So how do you see that sort of being, is this a one-size-fits-all or will they implement it differently? How do you see that?

Brian Vieaux (09:51)
I would say, I don't know if one-size-fits-all is the exact best description, but in effect it's built to be able to be implemented at companies of all sizes, shapes and models, including infrastructure around governance. So, I most recently came from a federally regulated financial institution, multi-channel. And as I thought back to the experience there in our own governance process, FRAME would have definitely worked in that environment.

And then as a correspondent buyer of loans for a lot of years, I worked with a lot of the correspondent sellers of all shapes and sizes, including the smallest shops, where oftentimes the chief risk officer might be the COO, in some cases the president. And the tool is built so that if a shop is set up like that, they could literally, they could use the tool to be able to at least demonstrate that they understand where their AI is or use cases are and how they manage, you know, how they're managing that risk. So, it is set up to be used at companies of all sizes. And we're already seeing some of the largest lenders tie FRAME into their framework and then they may have a more robust plan around it, but it's you know, FRAME becomes the foundation.

Marvin Stone (11:04)
So, for those of us who've been around MISMO for a while, there might be extensions to it, right? I mean that's kind of what we've seen in the data world is there's a data standard, but people will implement extensions, and it's really no different here with FRAME to go along the same way. But here's a question for you, you know, coming from the title industry. You know, I would like to know if a title agent receives this questionnaire from a lender, they're gonna get this FRAME document or FRAME questionnaire, how would they take it? Because you know, most title agents are never going to get a lender letter from one of the GSEs, right? So, Jonathan, how would you take that as you know, you've been a technology provider in the past, you get this thing. Where do you start?

Jonathan Kearns (11:47)
Well, I mean I would start with talking to my partner first and understanding what they're trying to accomplish, right? Again, we have in this industry, everything is on the onus of the lender to prove that stuff. So, they have to be able to understand how their partners are also doing it. And so that would be my first question. Then I would understand what they're doing and I'd fill out the information. If I'm not using AI as a title agent, cool. Then I could just say, hey, we don't have any AI use cases and products that we're doing it in.

But again, they are all, the lender is ultimately responsible for everything that happens and has the liability from a purchase aspect. I understand in the title industry, you know, on the notarization and things like that, there's a lot of liability there. But from selling it and having to repurchase a loan back, that's all on the lender.

Marvin Stone (12:36)
Yeah, for sure. So, plenty of incentive for everybody to get on board. So, Brian, just as we close up, any final thoughts on that and then Amy, I'm gonna ask you where people can find out more about FRAME.

Brian Vieaux (12:47)
Yeah, my final thought is I've been thinking a lot about the third-party providers that that make up the ecosystem, but I've also thought about the individuals, the humans in the organization. I could see a world where AI governance responsibility to some level will fall down to an individual licensee like a loan officer. And so, one of the things we encourage lenders as they're doing their inventory around AI use cases is to make sure that they're giving their loan officers in particular amnesty to bring all those use cases that the LOs might be using, the shadow AI, and make sure those get on that inventory sheet. So, engage your entire team, including your field sales force, the loan officers in this process.

Marvin Stone (13:31)
That's a great opportunity because loan officers certainly understand the requirements and the laws that they're under, but they don't necessarily understand everything about AI. And if they're, you know, just using a public model and they're training the model with you know NPI personal information, all kinds of things could go wrong there. So, Amy, for those who aren't as familiar with MISMO and MISMO being part of big part of MBA and then wanting to learn more about FRAME, where do they start? Where do they go?

Amy Moses (13:57)
Sure, yeah, you can go to the MISMO website, MISMO.org. We are a subsidiary of the Mortgage Bankers Association, and you can just search for FRAME in our search bar. And it will take you to a page where you can download it if you are a MISMO member. If you're not, you can contact me to learn more. And then there are three videos we've created about three of the primary artifacts that comprise FRAME. So, anybody can watch those to learn more about those specific resources and as you mentioned, there will be different iterations of this.

This is V1. And so, we are looking for feedback for those people who have already downloaded FRAME and started using it. There's a feedback form out there as well where they can provide us their feedback and then look for us to launch a series of webinars this fall that are based on the artifacts and the work we did at the workshop on Monday that you participated in, Marvin, that was very well received. People love getting into those small groups and really digging in to talk with each other about how they will apply FRAME. So, we want to carry on those conversations.

Marvin Stone (14:59)
Yeah, it was a great exercise for sure and there's a wealth of information out there already. As changes are made, you'll probably put that in the MISMO Matters newsletter?

Amy Moses (15:09)
We will, yes. MISMO Matters comes out monthly the third Wednesday of every month and we'll keep the community updated on any changes and additional resources as well.

Jonathan Kearns (15:17)
And while this was thought about for lenders implementing this, title agents can implement this as well. As long as they're MISMO members, and there's a lot of it's the mortgage industry, right? So, it's just not the lender industry, it's the entire ecosystem. And as you know, Marvin, because you lead a lot of the groups here, is that MISMO does a lot for the title industry as well. We partner with ALTA well, and so you know those title agents out there watching this, this could be something that'll help them. They're gonna need to understand their vendors. What are they, how are they utilizing AI? Because you guys have a lot of liability as well. And so, it's very helpful for them as well.

Marvin Stone (15:50)
And one of the things that we really didn't cover is it's just good business to know what your providers are doing with AI and there's a lot you can uncover with it. I'll just give you a couple for instances. one is we have a third party that we know of that just became aware of here at MISMO, who had a vendor and that vendor was not testing the model every two weeks or so and they found there was model drift. And so, you've gotta understand just to your point, Jonathan, it's not set and forget, this is work that's ongoing that needs to be done and needs to be monitored because as somebody said from the stage the other day AI is a different animal and it's not just deterministic like what we've had for years and years. So great.

Amy Moses (16:33)
And we just announced yesterday, Marvin, that we launched two new certifications and one is for advisory partners to help lenders implement FRAME if they do need that assist. And then another one is for vendors. So, to your point that you just mentioned, right, we'll have some controls. Jonathan worked on that certification and people can go to our website to learn more about that as a vendor. So, there's two new certifications where lenders can have more trust in their vendors and have people help them implement FRAME as well.

Marvin Stone (17:02)
Excellent. Okay, any final thoughts?

Jonathan Kearns (17:05)
No, just really appreciate you taking the time on this.

Marvin Stone (17:07)
Great. Thanks so much. Thanks for another episode of Stewart in the Studio and we'll catch you next time.

That's it for Stewart in the Studio. Find more episodes and insights at stewart.com slash lender. We'll see you next time.

Disclaimer (17:18)
This podcast is for informational purposes only and reflects the views of the speakers. It should not be considered legal or business advice, and listeners should consult their own advisors before making decisions.