Latest / Chrisman Commentary - Daily Mortgage News / 10.29.26 Fed Reaction; Sela's Nate Becker on Voice Agents; Economic Data Swath

Transcript

10.29.26 Fed Reaction; Sela's Nate Becker on Voice Agents; Economic Data Swath · Chrisman Commentary - Daily Mortgage News

• 0:02
  Speaker: Welcome to the Chrisman Commentary, Daily Mortgage News Podcast.
• 0:05
  Speaker: I'm your host, Robbie Chrisman.
• 0:07
  Speaker: Topics on today's episode include We have to talk about artificial intelligence, right?
• 0:13
  Speaker: Reaction to the Fed's rate hold from yesterday.
• 0:18
  Speaker: And my interview with Sela's Nate Becker on mortgage voice agents, their performance adoption, and the future role of loan officers.
• 0:24
  Speaker: Here, take a listen to a little preview.
• 0:26
  Speaker: Philosophically, if AI agents can demonstrably improve mortgage conversion and profitability, which parts of the broader mortgage sales and origination process are next to be automated?
• 0:38
  Speaker 1: I think that a lot of the like sales and process is going to be like is going to be augmented or it's going to be taken over by AI.
• 0:45
  Speaker 1: I think that, you know, it like in the next five years, I think that AI agents are going to be able to, you know, certainly the like origination process like process, I think that, you know, like building a loan file, those are things where uh, you know, it's really it's like it is, it is going to happen.
• 0:59
  Speaker 1: I think maybe uh the framing that I have in my head is, hey, what are the pieces that aren't going to be automated in the next five years?
• 1:05
  Speaker 1: And I think that those pieces are number one, I think the realtor relationship for distributed retail.
• 1:12
  Speaker 1: Uh, I think that it is, I don't think that realtors are going to be knocking down the door to hand their customers off to a process that doesn't touch a human at all.
• 1:19
  Speaker 1: And so, you know, I think that this is uh, I think that this is a piece where, you know, distributed retail has as quite an advantage here.
• 1:26
  Speaker 1: I think that they're also, you know, when we see this in the data, there's specific points where it isn't, you know, not in every um origination process, but in a lot of origination processes, there's there's a place where bringing in a loan officer can really provide some extra reassurance to a borrower, you know, and I think that like a good LO is really going to sort of lean into like where could they augment this sort of the backbone of an AI that can just sort of like run a process and run it along, you know, and and like, you know, and be patient and listen and you know, and then the the LO comes in to really sort of like augment at the right times and to make sure that everything stays on track and to kind of build that personal relationship.
• 2:03
  Speaker 1: And I think that that's going to be, I'd be surprised if there are AI agents that can't be that where like the lead to origination rate would not be benefited by an LO coming in at the right time, you know, to uh, you know, to sort of like add their their stamp of approval to, you know, and and their their their like you know, the level of trust.
• 2:25
  Speaker 1: And also just uh, you know, sit there with the with the borough and walk them through the what's you know, honestly, it's a it's a pretty scary process that you know where people have a lot of questions and you know and it can be stressful.
• 2:33
  Speaker 1: So I think that that's the piece that will, you know, at like five years out, still where top LOs are going to be they're gonna be originating maybe 10 or 20 times as many loans, um, and they're gonna spend their time doing this really, really high value add activity.
• 2:48
  Speaker: Thanks to today's podcast sponsor, FICO.
• 2:50
  Speaker: As the industry's most predicted credit score, FICO score 10T combines proven performance with deeper insight into borrower behavior to help support a stronger and more resilient housing finance system.
• 3:02
  Speaker: To learn more, visit FICO.com.
• 3:05
  Speaker: Artificial intelligence continues to capture the imagination of a wide range of companies, including lenders.
• 3:13
  Speaker: Why underwriting is the right problem for AI to solve first is a topic on a recent thought leadership piece on Chrisman Commentary, and it will be a topic on today's The Big Picture with James Says, the head of mortgage lending at Fifth Third Bank.
• 3:26
  Speaker: You can find all our shows on YouTube.com under Chrisman Commentary.
• 3:30
  Speaker: Check the live streams.
• 3:31
  Speaker: You may be thinking about rolling AI out, but who's in charge?
• 3:35
  Speaker: You or it.
• 3:36
  Speaker: In its IPO filing Monday, Anthropic warned prospective investors that artificial intelligence systems could pose catastrophic risks as they operate with greater autonomy.
• 3:46
  Speaker: Concerns have been raised both within our organization and by other researchers, regulators, policymakers, and industry participants that sufficiently advanced AI systems could, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures, pose catastrophic or existential risks to humanity, particularly if such systems were misused, replicated, weaponized, or allowed to operate with a high degree of autonomy.
• 4:10
  Speaker: And you want this to run your company?
• 4:14
  Speaker: Anyways, as expected, yesterday the Federal Open Market Committee voted ten to two to hold its benchmark federal funds rate in a range of three point seven five percent to four percent, managing the pace of tightening while preserving flexibility.
• 4:27
  Speaker: The committee is navigating a delicate balance between restoring inflation to the Fed's two percent target and avoiding unnecessary damage to resilient economic growth, AI-driven investment, and financial markets.
• 4:38
  Speaker: The decision produced an uneven market reaction.
• 4:41
  Speaker: U.S.
• 4:41
  Speaker: treasuries ended mixed due to oil-driven inflation concerns pushing yields higher, along with renewed Middle East conflict and expectations of a resilient labor market.
• 4:50
  Speaker: The FOMC's statement's emphasis on resilient economic activity, robust capital investment, and persistently elevated inflation appears to strengthen expectations for a potential December rate increase.
• 5:01
  Speaker: Although the Fed has offered no explicit commitment to further tightening, expectations for twenty-five basis point hikes in December and March suggest a gradual reversal of the 75 basis points of cuts implemented in 2025.
• 5:13
  Speaker: However, high rates may prove insufficient if energy-driven inflation persists, or elevated equity valuations continue supporting demand, and Walsh's rejection of conventional forward guidance has amplified market volatility.
• 5:27
  Speaker: The Fed's central challenge is maintaining inflation-fighting credibility without signaling a predetermined path that could unnecessarily destabilize financial conditions.
• 5:36
  Speaker: Warsh's longer term legacy will depend less on his communication strategy than on whether he fundamentally changes the Fed's balance sheet and liquidity framework.
• 5:45
  Speaker: His five task forces are laying the groundwork for reforms in communication, data, productivity, inflation, and balance sheet management.
• 5:53
  Speaker: Despite his preference for a smaller balance sheet, the Fed continues reinvesting principal repayments from its approximately two trillion at the T agency mortgage backed two trillion dollars of agency mortgage backed securities portfolio into treasury bills, preserving liquidity while gradually reducing its holdings.
• 6:12
  Speaker: Allowing mortgage backed securities to run off without reinvestment could withdraw a more direct source of support from mortgage markets and housing finance.
• 6:20
  Speaker: While changes to reserve management could tighten liquidity beyond what policy rate increases alone achieved.
• 6:26
  Speaker: Keep in mind that rate hikes tighten the price of money, whereas balance sheet reform can change its availability and distribution.
• 6:33
  Speaker: For today's interview, I wanted to welcome to the show Cela's Nate Becker to talk about mortgage voice agents, their performance adoption, and the future role of loan officers.io, has purpose-built agents that supercharge your borrower outreach, qualification, and origination across consumer direct and retail.
• 6:55
  Speaker: I do want to start with the basics a little bit when we talk about voice agents.
• 7:00
  Speaker: From your perspective, just where are we currently?
• 7:06
  Speaker: What are the capabilities?
• 7:07
  Speaker: I think people say, Yeah, I I don't know if I trust him.
• 7:10
  Speaker: Okay, I do trust him.
• 7:11
  Speaker: You're obviously originating tens of million dollars in mortgages daily.
• 7:16
  Speaker: Yeah, voice agents.
• 7:17
  Speaker: Thoughts on where we are currently, the capabilities, and then obviously things are evolving so rapidly, kind of where it's headed.
• 7:25
  Speaker 1: I think, you know, so first just a little bit about what we do.
• 7:27
  Speaker 1: Um, you know, at Sela, like uh we build highly polished, highly optimized AF voice agents.
• 7:33
  Speaker 1: Uh it's the, you know, for mortgage origination.
• 7:35
  Speaker 1: We're pure play just in mortgage origination.
• 7:37
  Speaker 1: And so we take bars through a large part of the you know mortgage sales process.
• 7:42
  Speaker 1: Um, we work with big consumer direct teams, we work with big broker shops, we work with large distributed retail teams.
• 7:47
  Speaker 1: And I think that maybe to answer your question, like at this point, I think that pretty much anyone can put together a voice AI that can say something over the phone.
• 7:54
  Speaker 1: And so I think the big question in everyone's minds, in everyone's minds, and I think the thing that is, you know, that we've been able to prove out over the last like 12 to 18 months is just that, you know, it's easy to build a voice AI agent, but the only thing that matters is are we actually driving like better, faster, cheaper experience for uh for borrowers?
• 8:12
  Speaker 1: And we've basically been able to, you know, we've been able to prove that at scale to directly answer your question.
• 8:18
  Speaker 1: There are large parts of the mortgage origination process, you know, where we run our uh like our AI agents head to head against human loan originators and in some circumstances against it head to head against other voice AI systems.
• 8:30
  Speaker 1: And we're able to measure the leads or origination rates for leads that you know that we work versus like leads or opportunities that these other teams work, and we're able to come out ahead.
• 8:40
  Speaker 1: And that's how we've been able to get successful, you know, that's sort of our bread and butter.
• 8:44
  Speaker 1: I think that our, you know, we have bots that can go as deep as pulling credit via an API.
• 8:50
  Speaker 1: We have bots that are triggering uh HELOC offer selection steps and then you know passing borrowers over to the rest of the work the rest of the app or do the the the rest of the uh the origination process themselves and then checking in if they you know if they fall off.
• 9:03
  Speaker 1: Like so we have we have bots now that go beyond just the you know the top of funnel, getting people on the phone, getting them interested, getting them, you know, like the first step to like towards qualified, and you know, and we can we can walk through large parts of the process.
• 9:16
  Speaker 1: And again, we've been able to do that and measure that and make sure that we do that as well or better than than human teams.
• 9:23
  Speaker 1: Um, so that's the it's a crazy world that we live in.
• 9:25
  Speaker: I talked to a lot of people on this podcast, and there's definitely a camp that says AI can automate everything in the mortgage process, and there's a camp that says, well, we're never going to replace humans completely.
• 9:37
  Speaker: I don't necessarily put a line in the sand.
• 9:39
  Speaker: I'm just here to ask questions.
• 9:40
  Speaker: So I will ask you for the sake of adding value, what does the data tell you about where AI agents are genuinely outperforming traditional mortgage sales teams?
• 9:50
  Speaker 1: Yeah, I think there are a couple of pieces.
• 9:53
  Speaker 1: I think this is a good question to ask, like, you know, where in the we're in the funnel and we're in the process, uh, you know, like AI agents tend to shine.
• 9:59
  Speaker 1: I think the first, the most obvious thing is that AI is just better at dialing and it's better at call timing.
• 10:03
  Speaker 1: Beats the lead, being persistent with follow-up, those sorts of things.
• 10:06
  Speaker 1: AI's just it's just got natural advantages over uh like a human uh human team.
• 10:10
  Speaker 1: And, you know, like whether it's a single, you know, like broker or uh, you know, a team of uh of you know 800 people, 800 LOs that are dialing out, AI is just better at this.
• 10:20
  Speaker 1: Um, we've also seen surprisingly large gains from using models to figure out the optimal next time to call each borrower.
• 10:27
  Speaker 1: So, you know, we think that like our dialing strategy is is has been pretty key to us being able to outperform.
• 10:33
  Speaker 1: I think that's pretty clear that that we just got an advantage over human teams in that regard.
• 10:38
  Speaker 1: I think there's also, you know, in the data, there's interestingly, there's a set of borrowers who seem to respond really, really well to AI.
• 10:45
  Speaker 1: Um, I think many of these borrowers, honestly, the average loan officer might not have infinite tolerance for them.
• 10:52
  Speaker 1: Um, so these borrowers tend to ask a lot of questions.
• 10:54
  Speaker 1: They tend to be the chatty type, they tend to, you know, sort of want to kind of take their time.
• 10:57
  Speaker 1: And I think that AI, one advantage is that it's able to be very patient and it's able to really, you know, kind of be willing to work with these borrowers and just go at their pace.
• 11:06
  Speaker 1: And we've seen that, you know, in this particular subset of borrowers that that we can really drive much better outcomes.
• 11:12
  Speaker 1: You know, the third big advantage is basically control.
• 11:14
  Speaker 1: Um, so, you know, our teams and by extension, our customers, they have a lot of visibility into what's actually going on.
• 11:20
  Speaker 1: Uh, that helps for compliance purposes.
• 11:22
  Speaker 1: You know, I think that our bots are more compliant than the uh than the typical LO, but you know, it also helps for performance.
• 11:28
  Speaker 1: Uh, we have much more fine-grained control.
• 11:30
  Speaker 1: For example, you know, if you want to run a promotion, it could just happen.
• 11:33
  Speaker 1: It's really hard to get, you know, every LO to pitch the promotion in the right way.
• 11:37
  Speaker 1: But this is something where it's just very easy to set up with AI.
• 11:40
  Speaker 1: And it also, relatedly, it also, you know, key to our success has been running many, many A-B tests or many, many, you know, experiments.
• 11:47
  Speaker 1: And it's so much easier to run experiments with an AI.
• 11:50
  Speaker 1: You just kind of give it two different sets of instructions and see which one performs better versus with a team of people.
• 11:54
  Speaker 1: It's it's a whole other thing.
• 11:56
  Speaker 1: So I'd say that those are the big things that stick out to me.
• 12:00
  Speaker 1: Um, I think there are a couple of pieces that were clearly human loan officers are just are are are substantially better.
• 12:06
  Speaker 1: And I'm not of the opinion that, you know, three years from now, five years from now, that this, that, you know, there aren't going to be any loan officers out there or that this thing is gonna be uh completely automated.
• 12:14
  Speaker 1: I think that the winning formula is gonna be really, really good LOs who can build really good relationships with consumers and they can come in at the right time, they can work their realtor relationships, you know, they can have the power of AI and the horsepower of AI and use it to augment the thing that you know that they're really good at, which is that, you know, that people piece that the that the AI is not necessarily, you know, isn't gonna be as good at.
• 12:34
  Speaker: Well, what do you see as the proper blend?
• 12:37
  Speaker: Because I'm I guess, or I would assume that we will get to the point where a loan officer will have an entered enough of their voice on recording that their AI agent will sound just like them and and that gets into a gray area morally.
• 12:50
  Speaker: But I thought what you said was really interesting about well, some of these borrowers are chatty and AI has infinite patience.
• 12:57
  Speaker: I had thought AI would be great for the conventional conforming vanilla loans, just speed it through the process.
• 13:03
  Speaker: I've done this before, blah, blah, blah.
• 13:04
  Speaker: But actually, you're saying it's the higher touch people that can benefit from this because they can they can take as much time as they need from an AI agent, whereas a human doesn't have it.
• 13:13
  Speaker: I thought that was very interesting.
• 13:16
  Speaker: I asked for some stats.
• 13:17
  Speaker: I have some stats here.
• 13:18
  Speaker: 40% higher profit than an existing call center from these agents, outperforming competitive voice AI platforms by about 41%.
• 13:25
  Speaker: Maybe I should have begun the podcast with this.
• 13:29
  Speaker: Five of the 10 largest independent mortgage banks now use sell it.
• 13:32
  Speaker: Congratulations.
• 13:33
  Speaker: That's obviously a huge deal.
• 13:35
  Speaker: What have you learned about the organizational, technological, trust barriers that determine whether lenders actually adopt AI agents at scale?
• 13:44
  Speaker: And AI agents is obviously the most cutting-edge wave of tech.
• 13:48
  Speaker: The last couple years, the mortgage industry, there were more vendors than lenders at a lot of these conferences, and lenders going, I don't, I have so much tech in my tech stock, I don't feel like I'm properly, I don't feel like I'm properly utilizing it.
• 14:00
  Speaker: How do I figure out where I can pair back?
• 14:02
  Speaker: I gotta, I gotta get rid of things.
• 14:03
  Speaker: So for you, getting getting over some of these barriers of adoption, but also when it comes to agents, because it's like, I don't know if I can trust an agent to go talk to borrowers.
• 14:12
  Speaker: There's obviously a lot of regulation in the mortgage space.
• 14:15
  Speaker 1: Yeah.
• 14:15
  Speaker 1: So I so first off, uh, you know, like thanks for thanks for the shout-out.
• 14:19
  Speaker 1: Uh, we actually just went live with our sixth top 10 IMB a couple of weeks ago.
• 14:23
  Speaker 1: So we're officially at six of the top 10 uh like independent merch banks in the in the country.
• 14:27
  Speaker 1: I think for me, the biggest learning that I have is AIs make really good demos, you know, and and I think that even before Voice AI was ready for prime time, it had good demos.
• 14:38
  Speaker 1: Um, I think the biggest thing that we've seen that has made us successful is just setting up clear quantitative success criteria.
• 14:45
  Speaker 1: So, you know, where is the ROI going to come from?
• 14:47
  Speaker 1: What can we expect, like given prior deployments, like you know, how many extra origin uh extra originations can we expect?
• 14:55
  Speaker 1: You know, like how does that compare with our current process?
• 14:57
  Speaker 1: How does that compare, you know, to the improvement that we could expect from working with uh working on some other use case first?
• 15:02
  Speaker 1: I think if the two sides do a good job of that, then we tend to set ourselves up for success.
• 15:07
  Speaker 1: And I think that where people have some sort of like general idea that they need AI, and we can't really land on, you know, a sort of like a business framework for um, you know, for how we're actually gonna, you know, like add to your to your top line, um, then you know, I think that that's uh that that that's that's where where it gets a little bit more challenging.
• 15:26
  Speaker: Obviously, agents will become more capable of handling more of the mortgage sales process.
• 15:31
  Speaker: What does the mortgage sales organization of the future look like?
• 15:35
  Speaker: And obviously that gets into the question of how fundamentally could AI change the role of human loan officers.
• 15:40
  Speaker: You mentioned kind of turbocharging them.
• 15:43
  Speaker: What is the what does the loan officer of the future look like?
• 15:45
  Speaker: What does the organization of the future look like, in your opinion?
• 15:48
  Speaker 1: Yeah, so it's interesting because, you know, in uh, you know, on like on one hand, like technology benefits, people can spend a lot on technology, and so that's the bigger companies, but I'm not sure that's actually how this is going to play out here.
• 15:59
  Speaker 1: So, you know, just to give you an idea, so you know, we operate at quite a large scale.
• 16:02
  Speaker 1: We do over 600,000 calls per day.
• 16:04
  Speaker 1: You know, you mentioned we work with six of the 10 largest IM uh IMBs out there.
• 16:08
  Speaker 1: Um, we, you know, we see almost 10,000 new lending tree leads a day, um, you know, which is just one of the lead providers.
• 16:14
  Speaker 1: So we're we're able to get to this scale that I think that previously, you know, it's really only the rockets and the lone deep bullets of the world were able to see this.
• 16:20
  Speaker 1: And we're able to basically take this, these skills that we've built, you know, over time, and we can kind of put it in the pocket of anyone.
• 16:28
  Speaker 1: Um, and so I think that that's a pretty interesting sort of change to uh like to the market.
• 16:35
  Speaker 1: And that sort of makes things, it makes technology kind of a leveler.
• 16:37
  Speaker 1: You know, we have this call center in a box.
• 16:39
  Speaker 1: And so, and now folks who are at much smaller scale, they now, you know, they they have the powers of the of these call centers.
• 16:46
  Speaker 1: And so I think that the best LOs out there are gonna take that, the power of what used to be the just the biggest consumer direct shops had this, and they're gonna be able to mix it with the retail kind of, you know, like the trust building, the like relationship building piece, and they're gonna be able to scale themselves a lot more.
• 17:02
  Speaker 1: So like take advantage of the scale that comes from the call center to box kind of thing, the volume that they can do, uh the flexibility, et cetera.
• 17:09
  Speaker 1: But then they really focus on relationship management, and that's that's a really killer combination.
• 17:14
  Speaker 1: You add that in plus AI, and I think that that's uh, you know, that's that's that's what that's what the future is gonna look like.
• 17:19
  Speaker: Certainly exciting times.
• 17:22
  Speaker: If you're looking for more information on the market leader AI voice agent, Cela.io.
• 17:28
  Speaker: Sela.io.
• 17:30
  Speaker 1: There you go.
• 17:31
  Speaker: Cool, man.
• 17:31
  Speaker: Really appreciate the time.
• 17:33
  Speaker 1: Yeah, it was great to meet you.
• 17:34
  Speaker 1: Um, thanks for having me on this.
• 17:38
  Speaker: Today's economic calendar kicked off with continuing jobless claims in at 1.7 million, about as expected, initial claims in at 194,000, just below last week.
• 17:49
  Speaker: Uh core PCE prices in at 2.5%, which is a little better than expected.
• 17:57
  Speaker: GDP in at 3.4%, a little better than expected.
• 18:01
  Speaker: Personal income month over month for September was up 0.2%, slightly missing expectations.
• 18:06
  Speaker: Personal spending for September was up 0.5%, slightly missing expectations.
• 18:11
  Speaker: Later today brings $44 billion of seven-year note treasury auction results.
• 18:17
  Speaker: And after the salvo of economic data, we find agency MBS prices a few ticks or 30 seconds better than Wednesday's close, the two-year yielding 4.70, and the 10-year yielding 5.02 after closing yesterday at 5.05%.
• 18:29
  Speaker: Let's wrap up with a joke and some housekeeping.
• 18:35
  Speaker: Last weekend I went to a Halloween party, and one of the other people who went there went as a former gifted student.
• 18:41
  Speaker: He just wore normal clothes, and when people asked him what he was supposed to be, he side and cried it was supposed to be a lot of things.
• 18:48
  Speaker: Thanks again to FICO for sponsoring this week's podcasts.
• 18:55
  Speaker: As the industry's most predictive credit score, the FICO score 10T combines proven performance with deeper insight into borrower behavior to help support a stronger and more resilient housing finance system.
• 19:05
  Speaker: To learn more, visit FICO.com.