Latest / Chrisman Commentary - Daily Mortgage News / 10.27.26 Origination Costs; Areal’s Argun Kilic on Automation; Second Guess the Fed

Transcript

10.27.26 Origination Costs; Areal’s Argun Kilic on Automation; Second Guess the Fed · Chrisman Commentary - Daily Mortgage News

• 0:02
  Speaker 1: Welcome to the Chrisman Commentary, Daily Mortgage News Podcast.
• 0:05
  Speaker 1: I'm your host, Robbie Chrisman.
• 0:07
  Speaker 1: Topics on today's episode include the cost to originate.
• 0:11
  Speaker 1: I feel like we talk about that every single day here.
• 0:14
  Speaker 1: Second guessing our Federal Reserve.
• 0:17
  Speaker 1: And my interview with Areal's Argun Kilic on automating closing disclosures, funding reviews, and post-closing tasks.
• 0:25
  Speaker 1: Here, take a listen to a little preview.
• 0:27
  Speaker 1: How do you see agentics evolving in the mortgage space?
• 0:30
  Speaker 1: Where does it go from here?
• 0:32
  Speaker: Yeah, that's a million-dollar question.
• 0:34
  Speaker: So trillion with a T dollar question.
• 0:36
  Speaker: Trillion dollar question.
• 0:38
  Speaker: Um look, I think Agentic is going to evolve into more of a reliable execution on your loans.
• 0:47
  Speaker: At this point, there are two types of approaches.
• 0:50
  Speaker: There are some companies that I see, you know, they're putting in a you know page-long prompt into resolution of one task, and God knows how often or how reliable that will operate when it is under stress and being utilized by hundreds of loans a month, right?
• 1:08
  Speaker: Or thousands of loans per month.
• 1:10
  Speaker: So I think the evolution would be as agents get smarter, uh, we would need to build solutions around how these agents can be accountable to what they do.
• 1:23
  Speaker: And you can, as if like you have a co-worker with you, you know, working, if something goes off, you can go back and look at the logs, if you will, or see what you know, you can ask, hey, what happened here?
• 1:35
  Speaker: I think we really need to have that accountability from the agentic executions in the market, especially in a highly regulated market where there's no room for any mistake.
• 1:46
  Speaker: I think reliable execution will become one of the most important things in the in this industry.
• 1:54
  Speaker 1: Thanks to today's podcast sponsor, FICO.
• 1:57
  Speaker 1: 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.
• 2:08
  Speaker 1: To learn more, visit FICO.com.
• 2:12
  Speaker 1: The big news today is that Microsoft is offering buyouts to around 60,000 employees to contain costs.
• 2:20
  Speaker 1: That's about a quarter of its global workforce.
• 2:23
  Speaker 1: Many are wondering if it portends what is going to come from AI in the coming months and years from other companies as well.
• 2:29
  Speaker 1: And on a slightly smaller scale, what's it going to do to the Seattle real estate market?
• 2:34
  Speaker 1: The ancient Chinese philosopher Sun Tzu said sometimes the best victory is the battle that was not fought.
• 2:41
  Speaker 1: In a shrinking residential market, there are a lot of potential battles.
• 2:44
  Speaker 1: For example, I'm hearing that some IMBs are not charging an origination fee or putting caps on LOCOM for jumbo loans.
• 2:53
  Speaker 1: It's especially true if the independent mortgage bank or broker is in an area where you're battling it out with credit unions and banks.
• 2:59
  Speaker 1: If you have a $2 million loan and an IMB's LO is charging a point, that's $20,000.
• 3:05
  Speaker 1: But if an originator at a credit union or depository bank is charging $5,000 for the same loan because they're capped at that, well, what's the borrower supposed to do?
• 3:14
  Speaker 1: So, yes, many IMBs do cap comp or origination fees.
• 3:18
  Speaker 1: Not only that, but with interest rates where they are and jumbo adjustable rate mortgages of viable products, often banks and credit unions have a solid offering.
• 3:27
  Speaker 1: There are some good deals to be had because of the poor affordability.
• 3:31
  Speaker 1: But buyers not in the business of buying homes have a hard time wrapping their heads around all of it versus five years ago.
• 3:38
  Speaker 1: There's only a small chance the Fed hikes this week, but an 80% chance they hike in December.
• 3:44
  Speaker 1: So many will say that this week's meeting is really about the December meeting.
• 3:47
  Speaker 1: But Warsha is going to try his hardest to avoid signaling a December move.
• 3:52
  Speaker 1: Why should he?
• 3:53
  Speaker 1: We'll get two more consumer price index and job reports ahead of that meeting.
• 3:57
  Speaker 1: And if we assume he successfully dodges those questions, the market will ask Does the statement suggest inflation is picking back up?
• 4:06
  Speaker 1: Or simply not returning to two percent fast enough?
• 4:08
  Speaker 1: How many dissent in favor of a hike?
• 4:11
  Speaker 1: The market expects three, so more or less could be a signal.
• 4:15
  Speaker 1: How do concerns over AI and events in France factor into Fed thinking?
• 4:19
  Speaker 1: A lot of questions.
• 4:20
  Speaker 1: I know.
• 4:21
  Speaker 1: The Fed will get a sneak peek of the personal consumption expenditure and gross domestic products, which are released Thursday.
• 4:29
  Speaker 1: So how does that info show up in the statement vote total?
• 4:32
  Speaker 1: If the FOMC actually expects to hike in December, why not just do it now?
• 4:36
  Speaker 1: If you expect to hike 75 basis points total, why not do it all at once and be done with it?
• 4:42
  Speaker 1: Plus, if rate decisions have a year lag, does a pause mean a pause this week mean the straight before moves remains closed for another year?
• 4:49
  Speaker 1: Those even more questions.
• 4:52
  Speaker 1: For today's interview, I wanted to welcome to the show Ariel's Argon Klitsch to talk about industry-specific artificial intelligence for the mortgage and title sectors.
• 5:05
  Speaker 1: He's identified the necessity of simplifying document-dependent processes through automation powered by natural language understanding, which led him to establish Ariel with the goal of providing dependable document understanding for mortgage and title documents.
• 5:18
  Speaker 1: I'm sure you hear this all the time.
• 5:29
  Speaker: So it actually arial stands for AI for real estate.
• 5:33
  Speaker: So it's Ariel, is what we call.
• 5:35
  Speaker: But I got this question a lot.
• 5:38
  Speaker 1: How did you determine the name?
• 5:39
  Speaker 1: Was it AI for real estate?
• 5:41
  Speaker 1: Is that where it came from?
• 5:42
  Speaker: Yeah, I mean, so the storyline is we basically wanted to have uh a name that's starting with A, uh for sure.
• 5:52
  Speaker: And you know, we looked at a couple of names, and you know, we picked for for example, we picked a name called Acuvian.
• 5:59
  Speaker: It's almost like a you know, a consulting complex name, but no one really liked it.
• 6:05
  Speaker: And ultimately we started thinking about like, and then we're thinking over overthinking about this, and uh, we decided to come up with a basic name.
• 6:13
  Speaker: Like, we're doing AI for real estate, so why not Ariel?
• 6:19
  Speaker: And that's the the beginning of it.
• 6:22
  Speaker: And um the other thing that I like about our name is our platform is designed to be extremely responsive, so there's also a real time factor to it.
• 6:33
  Speaker: So everything came together.
• 6:34
  Speaker: So AI for real estate in real time.
• 6:36
  Speaker: That's what Ariel stands for.
• 6:39
  Speaker 1: Very cool.
• 6:39
  Speaker 1: Well, what does it actually do?
• 6:41
  Speaker 1: And what was your, or maybe I should say, what was your goal when launching the company?
• 6:45
  Speaker 1: But then also, yes, take us into the product.
• 6:47
  Speaker: At the highest level, we have built AI automation solution for the mortgage operations.
• 6:53
  Speaker: Uh, we have two flagship products, CD Balancer and uh Copiled Agent.
• 6:59
  Speaker: Um, CD Balancer entirely automates the C to CD balancing process for lenders.
• 7:05
  Speaker: We in fact released the generation for a couple of weeks ago.
• 7:08
  Speaker: And the Copiled Agent is the Agentic platforms to get the tasks done for processes like funding review, post-closing review, and for other off-stream tasks.
• 7:18
  Speaker: And the starting point of Ariel was going back to 2019.
• 7:24
  Speaker: You know, I was looking into the mortgage market, and there was no proper OCR solution.
• 7:30
  Speaker: The best solution in the market was at 50% accuracy.
• 7:33
  Speaker: And two years prior to that, Google announced this Transformers uh paper, uh Attach is all you need.
• 7:40
  Speaker: So I was thinking about like, can we apply the new concepts, new AI concepts in the mortgage world and build a new document AI layer uh that can provide reliable data?
• 7:52
  Speaker: So that's how we started a company.
• 7:53
  Speaker: First couple of years, we purely spent our time on improving the document AI for um mortgage.
• 8:01
  Speaker: So we built this entirely from ground up for uh mortgage.
• 8:05
  Speaker: And as we get into what 2023, one of our first product managers was uh a closer, and he came in and said, Hey, closer stack needs AI.
• 8:16
  Speaker: Can you guys build CD balancing solution?
• 8:18
  Speaker: And we built that in 2024.
• 8:21
  Speaker: Now we thought about like, what about the closing reviews?
• 8:24
  Speaker: Can we do something about that?
• 8:25
  Speaker: We launched the closer co-pilot, and as Agentic AI evolved in, we basically, why are we bonding ourselves to just the closing?
• 8:34
  Speaker: Why we can do pretty much like we can build agents for different types of processes, so it has evolved into copilot agent.
• 8:42
  Speaker: So that's historic area.
• 8:44
  Speaker 1: Copilot agent, obviously, agentics is a very hot topic in the mortgage industry right now.
• 8:50
  Speaker 1: What is the latest with agentics in the industry?
• 8:53
  Speaker 1: Maybe what is what is co-pilot agent accomplish, but at large thoughts on where we are with the mortgage industry when it comes to agentics?
• 9:02
  Speaker: Yeah, absolutely.
• 9:03
  Speaker: So I think we have two types of companies in the market or two types of approaches with the Agentic AI.
• 9:12
  Speaker: The one of which is basically using agents to operate on a reliable data set to achieve something that you can trace.
• 9:24
  Speaker: The second type of agentic evolution is right now happening, is some of the newer tech companies try to come in and write a you know one page long prompt to automate a task.
• 9:40
  Speaker: The problem with that is that it is not reliable and consistent.
• 9:44
  Speaker: If you have an operation with 500 loans that you'll close every month, you want to make sure that the agents actually operate reliably.
• 9:52
  Speaker: So, going back to your initial question, what is Copilot Agent?
• 9:56
  Speaker: It's actually a reliable agentic automation platform for mortgage processes.
• 10:00
  Speaker: What we have done is you create an agent in the runtime and add a task to it, and it builds an execution for you which operates and executes consistently across all the loans that you have.
• 10:16
  Speaker: And you can audit and trace what happened and how um decisions are made towards closing a task or spinning off an exception for your closures to review.
• 10:28
  Speaker: With this reliable approach, companies, companies working with us uh were able to get to a real automation.
• 10:36
  Speaker: For example, and they're seeing great results.
• 10:38
  Speaker: And for example, one of the top-down lenders we work with, they see 98.8% task automation.
• 10:46
  Speaker: So, in other words, if you have 100 tasks that you need to review, 98, almost 99 of them will be auto-completed by Ariel.
• 10:57
  Speaker: That leads to a number that we trace, which is a trueput.
• 11:01
  Speaker: And from our perspective, true put means the number of loans that you don't need to touch.
• 11:06
  Speaker: Now, lenders that we work with, roughly about eight out of 10 loans, they don't need to even look at it anymore for uh when it comes to funding or post-closing review.
• 11:15
  Speaker: So that's what CopilAgent is.
• 11:17
  Speaker: And in fact, we filed a patent and it's pending it patent right now.
• 11:22
  Speaker: We're really proud the architecture that we have built so that we can deliver a reliable execution solution for our um customers.
• 11:29
  Speaker 1: Are you still encountering resistance to your product at this point?
• 11:34
  Speaker: And if so, how or where?
• 11:36
  Speaker: That's an excellent question.
• 11:38
  Speaker: So um, look, we've been in the market for about six years.
• 11:41
  Speaker: And in the first three to four years, the resistance was about can I rely on document AI?
• 11:48
  Speaker: Uh we have proven multiple times that we can deliver reliable data.
• 11:54
  Speaker: So, for example, you're running at 99% accuracy for critical data fields to be parsed from a document.
• 11:59
  Speaker: So that resistance is almost gone because lenders onboarding with us see that immediately.
• 12:05
  Speaker: But the resistance that we're seeing right now is going back to can I rely on the agentic executions?
• 12:13
  Speaker: But copiled agents showing how you can trace and audit the results, and it actually operates consistently across all of these loans addresses that problem.
• 12:24
  Speaker: And that's why, you know, initially after onboarding, trying a couple of loans, lenders actually quickly see that it is reliable.
• 12:32
  Speaker: So that resistance is also with us, especially uh, you know, gets addressed very quickly and they actually feel confident about the operations.
• 12:42
  Speaker 1: Companies are inundated with this is the latest and greatest.
• 12:46
  Speaker 1: This is the latest, and it's hard for them to actually figure out which one to choose, how to properly implement it, how to train up their staff to maximize utilization, how to figure out the ROI on it.
• 12:57
  Speaker 1: And so I think that there is like this this pushback recently of AI was supposed to lower the cost to close alone.
• 13:05
  Speaker 1: Well, we haven't seen the cost to close alone come down.
• 13:07
  Speaker 1: So should we be rethinking our AI strategy?
• 13:09
  Speaker 1: That's that's what I've heard a lot of recently.
• 13:12
  Speaker: I mean, that's an excellent question.
• 13:14
  Speaker: So um going back to our experience, right?
• 13:18
  Speaker: I mean, the initial years, it was about like, hey, can I rely on document?
• 13:23
  Speaker: And if you end up reviewing every document that we process, you cannot obviously save money, right?
• 13:29
  Speaker: We are past beyond that.
• 13:32
  Speaker: But then it really comes down to a point of right now, can I can I really do I have to review every agentic execution?
• 13:40
  Speaker: Again, you won't be able to save money.
• 13:42
  Speaker: So that is why as a company, we we spend a lot of time on our product to figure out how we can hand off the knowledge to the operator, saying that I have done X percent of your tasks and you have Y percent left.
• 14:00
  Speaker: And if you can build that operational handoff, then your closers, funders will be able to rely on your solution and do less work and increase their efficiency.
• 14:13
  Speaker: So that is why we are tracing these throughput numbers, task completion numbers that I mentioned earlier, to ensure that ROI is there.
• 14:21
  Speaker: And as lenders see that, wow, 98 out of 100 tasks are automated.
• 14:26
  Speaker: Sure, there is an ROI, sure, there is an efficiency gain, sure there is uh a path forward in terms of if uh you know scaling your organization.
• 14:36
  Speaker 1: Well, you're obviously trusted by some incredibly well-known companies.
• 14:40
  Speaker 1: I was on your website and it doesn't take much scrolling to see that you're working with rate, old republic, notary camp.
• 14:49
  Speaker 1: These are big companies in the industry.
• 14:51
  Speaker 1: What was the process like to get approved with all of them?
• 14:55
  Speaker: The process really started with the product.
• 14:58
  Speaker: They really liked uh in each different case, we they're actually utilizing a different product in our stack.
• 15:06
  Speaker: And they always liked what we're trying to do, how we approach the efficiency gains, how we tie this back to their operations to ensure that the handoff will be clean and they can rely on this efficiency gain.
• 15:23
  Speaker: Some of these processes took longer than the others, but it really coming back to the product.
• 15:27
  Speaker: We show the product over time.
• 15:30
  Speaker: They basically some of them had feedback like, hey, can you do XYZ?
• 15:34
  Speaker: And we are a nimble company.
• 15:35
  Speaker: We are a really nimble and fast company.
• 15:37
  Speaker: We actually listened to our customers really well.
• 15:40
  Speaker: We applied their feedback and recommendations into our solution.
• 15:43
  Speaker: Over time, this turned into a partnership.
• 15:45
  Speaker: And then they started working with us.
• 15:47
  Speaker: We're really, really proud of them.
• 15:48
  Speaker: Um, obviously now we can reference to how we operate with these large customers in the market, and that is a uh a bigger signal.
• 15:57
  Speaker: Our next customers can actually see what's happening with the other bigger clients that we work with, and they get really excited.
• 16:04
  Speaker: But I also want to open up Dip Note.
• 16:06
  Speaker: I mean, some of the customers that we work with, Guarantee Rate, et cetera, they're also great partners.
• 16:11
  Speaker: Uh, we really have great partners.
• 16:12
  Speaker: We don't really necessarily call them our customers.
• 16:16
  Speaker: Honestly, we really, really welcome and you know, enjoy working with our customers at this point.
• 16:21
  Speaker: But to a point, you know, product is a starting point.
• 16:25
  Speaker: And obviously, following that, we meet you know, high standards of security, we meet the high standards of the compliance, like you know, InfoSec checks and et cetera.
• 16:35
  Speaker: Uh, we go through those processes and ultimately we close up these uh partnerships again with the product.
• 16:42
  Speaker 1: What's your roadmap going forward?
• 16:44
  Speaker 1: What's next for you?
• 16:47
  Speaker: We always say that uh to our customers, uh, you haven't seen anything yet.
• 16:51
  Speaker: Um honestly, we are building the future of the mortgage operations.
• 16:57
  Speaker: Historically, we have always been focused on the closing side of the operations, right?
• 17:02
  Speaker: Anything starting with post underwriting, you're ready to go and close alone, and all the operations under that process cycle we address.
• 17:12
  Speaker: Moving forward, we will expand into a few other areas and actually we'll make a couple of more announcements.
• 17:18
  Speaker: But yeah, we will offer more uh product scope to our customers across the entire lifecycle of the mortgage.
• 17:25
  Speaker: Um, for example, we have a solution for the processors today, like copiled agent upstream actual helps processors, but there's some room in between the processing and closing that we can address.
• 17:38
  Speaker: So that's that's the next steps that we'll be waiting for Ariel.
• 17:42
  Speaker 1: Certainly exciting.
• 17:43
  Speaker 1: I know you have a lot on your plate and you're very busy, so I appreciate you making time for me.
• 17:48
  Speaker 1: The website is ariel.ai.
• 17:50
  Speaker 1: Go check it out, people.
• 17:52
  Speaker 1: And uh hopefully we'll talk again soon, Mr.
• 17:53
  Speaker 1: Klitsch.
• 17:54
  Speaker: Thanks so much.
• 17:55
  Speaker: I appreciate the time.
• 17:56
  Speaker: Robbie, uh, thanks for hosting us.
• 17:58
  Speaker: And uh I really enjoyed the questions and the conversation.
• 18:00
  Speaker: Thank you.
• 18:03
  Speaker 1: In terms of intraday bond market activity and therefore interest rates, U.S.
• 18:07
  Speaker 1: Treasuries and agency mortgage-backed securities traded mixed but largely range-bound yesterday, ahead of this week's October FOMC decision, with shorter maturities underperforming, a mid-a-week two-year note auction, and the long end posting modest gains, while easing oil prices and a rebound in French government bonds provided only temporary support.
• 18:25
  Speaker 1: Investors must balance potentially attractive mortgage-backed security valuations against spread volatility, refinancing uncertainty, continued housing weakness, and the possibility of renewed inflation-driven rate pressure.
• 18:37
  Speaker 1: Agency mortgage-backed securities posted their strongest weekly excess return performance since August last week, supported by safe haven flows and relatively contained volatility, with conventional and Ginny May 30-year securities appearing attractive on a spread basis, particularly seasoned mid-coupon pools.
• 18:53
  Speaker 1: Today's economic calendar includes updates on September durable orders, August FHFA housing price index, August SPK Schiller Home Price Index, October Consumer Confidence, September New Home Sales, and also includes a $70 billion five-year treasury note auction.
• 19:08
  Speaker 1: We begin the day with Agent CMBS prices slightly better than Monday's close, the two-year yielding 4.69, and the 10-year yielding 4.97 after closing yesterday at 4.99%.
• 19:18
  Speaker 1: Let's wrap up with a joke and some housekeeping.
• 19:23
  Speaker 1: Last year I dressed up as a screwdriver for Halloween.
• 19:26
  Speaker 1: It wasn't the best costume, but I still turned a lot of heads.
• 19:30
  Speaker 1: 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.
• 19:48
  Speaker 1: To learn more, visit FICO.com.