Rates drop, you hire. Rates rise, you cut. It's been the same playbook for decades, and it still costs $11,000 to originate a single loan. AI agents are starting to look like the way out. But the pressure to move faster, do more, and stop adding headcount isn't just something lenders are feeling. It's the same intensity Nima Ghamsari, Co-founder and CEO of Blend, has been driving inside the company, and it's shaping how the industry's best operators are starting to think.
In this National Mortgage News Leaders episode, Michael Moeser sits down with Nima to go beyond the hype and get specific: what does an agent-first operation actually look like, where does the industry stand today, and what should executive leaders be building toward right now?
Tune in for:
• What the AI potential was two years ago and how it's perceived today
• Why "agent first" beats "AI native" as a model for mortgage operations
• The production capacity and accelerated decision-making lenders can expect
• Strategies to capture more opportunities and prioritize high-intent applicants
• How AI-driven workflows with human review support compliance and consistency
Transcription:
Transcripts are generated using a combination of speech recognition software and human transcribers, and may contain errors. Please check the corresponding audio for the authoritative record.
Michael Moeser (00:09):
Hello everyone, and thank you for joining us. I'm Michael Moser, senior content strategist at National Mortgage News and I'll be your host for today's leaders episode. I had the pleasure of interviewing today's guest, Nima Ghamsari, co-founder and CEO of Blend almost two years ago and I'm really looking forward to expanding the AI and mortgage conversation. Today we're going to revisit some of Nima's predictions from 2024, explore the mortgage industry's growing capacity problem and discuss why an agent first approach may be the way forward to solving the industry's age old capacity problem. Welcome, Nima.
Nima Ghamsari (00:47):
Yeah, thanks for having me.
Michael Moeser (00:49):
Nima, in our earlier interview, we learned a lot about your background as a professional poker player, paying his way through college and how it helped you in your professional career to better read the room. Whereas I would say in this case, read the market to more clearly identify trends, opportunities to pursue and pitfalls to avoid. We're going to be asking you to look back at some of those predictions and then looking forward to where things are headed. But before we get into the questions, can you level set us on where the mortgage industry stands in mid 2026?
Nima Ghamsari (01:22):
Yeah, I'd say again, thanks for having me. Mortgage has an immense opportunity right now and I'd say even banking more broadly has an immense opportunity because so much of the work in mortgage and more broadly with banking is around reviewing documents, reviewing paperwork, and the cost of a mortgage are $11,800 in 2025 as of 2025, which is up 35% in three years, it's growing, not shrinking, even as we get all this AI, this magical AI technology being created in the background. And so there's fewer people doing the same manual work because the mortgage volumes are down, but they're doing the same manual work and that just means a higher cost per loan. I think there's also a conversion problem where there's so much work that has to get done in every file that things get lost in the shuffle. I hear this every day with our customers that somebody asks the customer for the same thing two, three, four times and that adds the cost, that adds the frustration.
(02:24):
And so as volume starts to grow again and it started to grow again in the first quarter of this year, the industry can't live in this world of hiring and firing like we had to in the last boom and bust. And so the only answer in my mind is to rebuild how work gets done and have agents take a first pass at everything, replacing a lot of the manual touches and then sending it to a human for that review. So the agents doing the document review, the calculations, et cetera, and sending it to humans to do the final review and make sure everything was ticked and tied. And that brings the cost down, that makes the workforce really scalable and still keeps humans at the centerpiece of getting this really important transaction done.
Michael Moeser (03:03):
Well, thanks for level setting us there. And I think that's a great segue to our conversation back in 2024. When we first spoke, you talked about exception handling and document verification as early AI wins lenders should consider deploying. Which lending workflows have proven their potential and become table stakes since then?
Nima Ghamsari (03:26):
Well, it's interesting. I would say in 2025, so right after our interview, there was a lot of hesitation around AI. There was worries about what would happen with the data. There was worries. There were worries about hallucination. And what we've seen is this dramatic improvement in what capabilities happened. And the concept of agents was brand new at that time. And now the agents, which I think of as a scalable workforce, you spin one up, new task comes in, you spin one up, it does some work, you spin it back down. That has become a really mainstream concept that actually works.
(04:02):
And so what used to have to happen was you had to create in an old rules-based world, you had to create a happy path and say, "I'm going to be able to solve for these 40 or 50% really fast, fast lane, and these other 50 or 60%, they're too complicated for computers and that's where my humans are going to spend the time working on it. " I think agentic AI really changes the nature of that. The agents don't care about how hard the math is or what kind of document came in or how long the guidelines are or how much the guidelines have changed or even what product it is. It doesn't just apply to mortgage, it applies to everything that a bank does agents just care about what is the task at hand, what are my instructions, what is the context that I have?
(04:44):
And they can do just about anything. So I think we can move from a world of thinking about exception processing to a world of a truly elastic workforce that's at your service, at the human service of I have a lot of work to get done. I need to spin up 30 agents to go work on this. And that means so much for our customers, the banks and lenders in the marketplace because they've been dealing with this boom and bust for decades.
Michael Moeser (05:11):
Well, that leads into my next question because back in 2024, you also talked about starting with a very focused limited set of objectives with AI initiatives to achieve early wins, get ROIs, scores on the board, if you will, which have continued to prove elusive for many organizations in terms of getting those ROIs, even outside of financial services. Would you still say that's the best approach given all of these new advancements such as agents? How does your advice change today?
Nima Ghamsari (05:44):
Well, and I actually see this a lot in the market too, where people will take a very narrow use case and I'll give a hypothetical example, appraisal underwriting or fraud review for a new customer or member coming in and they'll say, "Let's go and build something that can handle fraud or appraisals." And I think that was a great strategy to dip your toe in the water in 2025.
(06:09):
But what's happening now is because the agents are getting more powerful and they're able to do more tasks, the amount of work you can give them, if you build the right kind of system is pretty much any of that. You have to give it access to the right tools and the right data and the right instructions and it can do pretty much anything you want on demand. And so I'd say a lot of our focus this year at Blend has been how do we give it access to those tools and those data points and the documents that are coming in as they come in in real time. I like to describe this as ambient intelligence. This is living in the background. This doesn't wait for a human to click a button like you would in ChatGPT. This relies on a human, you could be asleep, a human to come in afterwards and review the work that was done by the agent.
(06:56):
So a lot of my focus now is how do I make this something that's generalized and it can handle and understand and truly create that scalable workforce for our customers. So it doesn't matter what event comes in, like I said, it could be a new document, could be an appraisal comes in, could be a fraud report, could be a credit report, could be just new data, the consumer updates there, the product that they're interested in. Well, that changes everything. That changes the whole product situation, changes everything you have to underwrite,
(07:24):
But agents are capable of that today. And so what I focused on is making that possible, but then also making it as turnkey as possible for our customers to turn on. So the nice thing that we have at Blend is we have this harness, which is all these events coming in, new documents, new data from the consumer, from third parties, gift letters, appraisal, title, et cetera, coming into our system. And then we have these output tools of, I need to request something new, I need to do a calculation, I need to update a data field. So we have all those things already built into our system. We've been trying to make this turnkey for our customers where they to turn on autopilot, which is our flagship AI agent, just check one box in our system and our configuration and now you have an army of agents working for you and that is really cool.
(08:09):
And any new data that comes in or will trigger the autopilot agent to wake up, spin up that capacity, do some work and then push it back to the system for more information.
Michael Moeser (08:20):
I'm imagining you with a megaphone saying unleash the agents and they're tackling work. But as we know in the mortgage industry and then financial services, not all challenges are equal. Are there any mortgage AI use cases that have turned out or are still turning out to be harder to implement than the industry expected and why? And I guess for the use cases mentioned, are the challenges or barriers preventing their success temporary and/or easily surmountable? Or are we talking some systemic issues that require significant changes within mortgages itself?
Nima Ghamsari (08:59):
Yeah. I mean, I'd say one thing that's almost accepted truth at this point is that things will evolve quickly and so anything that's not possible today we should assume will be possible tomorrow.
(09:15):
So just to cover off on some stats for our customer base right now, of the several hundred customers that Blend has, lenders and banks, 65, we've been live in market with Autopilot for a month and a half because we made it so easy, 65 have already turned on AI agents and almost 10,000 loan applications already and it's driven the cost down of doing those. We've seen cycle time for our customers who have turned it on for their loans go from 29 days to 21 days. So it's real numbers, real improvement because we've made it easy for them to adopt. By the way, a year ago when people were skeptical of this, of AI and Agentic AI and wondering if they should do that in 2027 or 2028, I'd say that mindset has shifted. I'm seeing our customers look at this capability and say, it's both an existential risk if I don't do it, but it's such an opportunity for us if we do it.
(10:14):
And they think of speed now. How fast can I adopt things as a moat for them? It came up yesterday. I was on site with a fairly large partner of ours and we talked about speed being a moat. How can you be faster at adopting these things than your peers? Now, that doesn't mean it's perfect that every use case. So to get back to your original question, there are some really complex compound use cases that are very good, very important, but not trivial to solve today. I'll give one example.
(10:45):
If I'm working with a lender and I'm your customer, you're my loan officer, I'm your customer, you see that I have bonus income. Well, as a loan officer, you'd say, "Oh, I need to get end of year pay stubs to make sure that your bonus income is consistent over the past couple years." And then that triggers a whole bunch of downstream things. Well, if the income changes as a result of my end of it ... So one, you have to calculate, you have to read the documents, you have to calculate the income, you have to compare it to what's on the application, you have to update the application. If the income goes down though, that triggers a whole bunch of downstream things. Now, I don't think that's perfected today in autopilot or really any system. It's currently done by humans going through those downstream chain of things like your DTI may change, so you have to rerun the underwriting system.
(11:26):
If your income changes, there's thresholds for what are considered large deposits, which may be loans that you have to go and look through in your bank statements, then you have to read every single bank statement. And when you run the underwriting system, it might come back with additional conditions that you have to request of consumers. And so one end of your pay stub can trigger a whole chain of events. And so this is something that we're really excited about and we're building and I think it's not a well-solved problem today. I'd say the other part that I'm excited about is around the conversion problem that I talked about earlier. How do we make sure that the consumer is getting a hyper-personalized view of what the right financial outcome is for them in that flow and give the loan officer the tools to give that consumer that hyper-personalized view, which a very experienced loan officer may be able to do, but someone who's only been in the industry and we talk, one thing I think about a lot is that there's a lot of turnover in the industry.
(12:24):
In some of these two or three years in the industry, they don't know the 50 loan programs that might be available at a bank or a lender and they don't know how if DTI goes from 43 to 45, what that may affect, or if LTV goes from ... They don't even know what LTV is, but they don't know in some cases, but if they don't know if LTV goes from 68 to 72, that may change the rate. So those are real considerations that I think drive conversion because if you can show people the best financial situation for them, they're more likely to want to transact with you, also make it easier for them to work with you. So that's another big problem that I'm excited about. And I think while we have some early things in the works there, we're not out yet inside the autopilot framework.
Michael Moeser (13:06):
Sounds like you're learning to adapt as and adapt the model on the agent along the course of the way. Now you also talked about solving issues in real time instead of sending loans to exception queues for additional review. Why is this especially important today as the mortgage industry faces a capacity problem? How close are lenders today to reaching that true real-time mortgage workflow?
Nima Ghamsari (13:33):
I believe, and we have two pretty large customers who are doing this with us right now, where they are going to be able to have a consumer go from the initial touchpoint to see their offer, to get approved from their offer for their offer, lock their rate, and prepare for closing, and then go through the fulfillment steps in a self-driving manner. The reason we call it autopilot is because it has to be self-driving, but go through those fulfillment steps in a self-driving manner until they're ready to close, until they're clear to close. So I think we're going to have that very soon. In the next couple months here, we're going to have that finally exist in the industry. It's something that we've all wanted forever, but is practically very difficult because there's thousands of pages of guidelines and there's so many different product types and so many different consumer situations.
(14:22):
Now, but I don't want that to be misconstrued. What that doesn't mean, and one thing we're building into there is as they're going through that sort of self-driving funnel, doesn't mean that the consumer doesn't want human help. And so not even exceptions, but even just how do you help open the door? So how do you make it really easy for a consumer when they get to that page, when they're looking at the product that you're recommending to them, whether it's on the phone or just online, that they can get in touch with the right person, the right loan officer to help walk them through that situation. So we're not trying to dehumanize the process. We're trying to take the manual reading of documents, reading of data, moving to data and documents between systems. I mean, there's been this word, this phrase, stare and compare in the industry forever.
(15:09):
That still happens. And check the checker, that still happens. It doesn't have to happen anymore. So that's the stuff we're trying to get out of the way so the consumer can work with the lender in the way that makes sense to them. So that's one. And the second thing, I still think at this point in where we are with AI, this may change over time.
Michael Moeser (15:26):
Okay
Nima Ghamsari (15:26):
The final review, the lender should take a final review of all the work that AI has done. And so one of the very nice things about agents is that they will document every single step that they take along the way. And that means that everything is perfectly auditable, it's fully understandable, it links back to the document or the guideline or whatever that it read. And so humans should still take that final review because a lot of money is changing hands and we have to make sure that when that money changes hands, that there's real confidence that the real person with real data and the correct math and all those things matter. And so we're not saying take that away either, but we are saying make it so the technology gets the really mundane stuff out of the way so that the consumers can focus on what they need, the lenders and loan officers can focus on what they need and the system just works with lower friction.
Michael Moeser (16:18):
So where are the real potential gains to generate capacity in the lending process? You talked about the capacity crunch, you've talked about the challenges with turnover. I mean, is it in the marketing qualification step? Is it the application, underwriting, funding, another segment? I'd hate to say all of the above, but curious, where are the real potential gains?
Nima Ghamsari (16:43):
Well, I think the unfortunate thing with any industry that's historically been this much reliant on humans for both the value add work and the less value add work that is required is, but the real challenge is if mortgage volumes go down a percentage point and now your mortgage volume as a lender doubles, it's almost impossible to serve all of the two times the capacity with the same people. And so unfortunately, I think the bottlenecks are probably everywhere in the funnel because there's $11,800 in cost and a lot of that is human. A lot of that is humans. And so you can't just pick one area. Almost all of the areas require some amount of human intervention. And so I would say that my hope is that we can turn the humans in this process into superhumans and make them able to have these, again, this army of AI agents working for them and being able to handle 10 times the capacity.
(17:53):
So as new information's coming in and AI agents looking at it in the background, this concept of ambient intelligence, doing the work for them, and then serving them up a really nice summary of, here's all the stuff, here's the things that you should pay attention to and
(18:08):
Here's the next steps that we think you should take on this file. And now, of course, the human might have their own view. A really smart loan officer might have a better idea than an AI. A really smart underwriter might catch something that even a really good AI might not catch or think of something that a really good AI may not catch, but that doesn't mean that the humans should be going and doing the stare and compare and the check the checker. That kind of work that slows things down today and is a bottleneck needs to go away. And every lender, it's interesting, every lender that I ask about the state of the industry, they all agree that in five years humans are not going to be manually reviewing 100% of documents multiple times per loan like we use today. And I think it's basically they're not going to be reviewing documents at all anymore because the AI agents are going to do the document lunging and the calculations and the moving point A data from point A to point B, and they're going to be doing all of that.
(19:04):
Agents are going to be doing all of that. So they all agree with that. And the question is how do we get from here to there?
Michael Moeser (19:10):
Well, let me ask you a question in our earlier conversation. There was a distinction that we made in terms of agent first versus AI native. And can you really explain for the audience the difference and why agent first is a stronger model for mortgage origination operations? What are the benefits for both lenders and borrowers?
Nima Ghamsari (19:32):
Well, when I think of AI native, when I think of myself using AI, I think of going into ChatGPT or Claude, entering a prompt, getting a response, probably the right answer, or at least a good answer, but still me being the centerpiece, me having it to navigate and click the buttons to do that. And so it's like being in the cockpit of a plane that has really good telemetry that has a lot of things automatically done for you, but you're still actually the one moving the steering wheel up and down and left and right. But I think agent first is a little different or a lot different than that and that the humans are not triggering the agents to work. The humans are not clicking a button. In fact, they're not clicking any buttons. The humans are not going into ChatGPT and typing something. The work is all being orchestrated in the background with this ambient intelligence and it's done, like I said, new event comes in, you spin up an agent, not you, the system spins up an agent that does the work and then that could be at two in the morning and that's the unlock.
(20:43):
If things can happen in real time, 24 hours a day, seven days a week,
(20:48):
And that can happen at two in the morning, consumer just finished a long day at work, they have a late shift, they get in their mortgage application process, they're applying for this mortgage, they're uploading documents, all this work is happening, they're choosing what they want. And then the loan officer wakes up and the agent first model is more of the loan officer wakes up 8:00 AM the next day, sees that file and it's already been sort of fully thought through. The file has been dynamically and progressively underwritten so the consumer, they have a trust that gets pulled in. If they have bonus income, like I said, that gets pulled in, that pulls other things in downstream and the human is like, "Oh, great. Well, this package has already been nicely put together for me. I'm going to give this person a call and help them walk through the final steps of the process." And so the difference between AI native is ... And so I guess the agent first, the way I think about it is now your air traffic control, you're not in the cockpit of the plane.
(21:37):
You have all these planes flying around and your job is to make
Michael Moeser (21:41):
Sure
Nima Ghamsari (21:41):
That you land all of them. And so you're sitting in a totally different tower far away and it just happens that you're controlling all these planes. And so the difference is, again, does a human have to initiate the action or does the action happen and then it comes to the human as a secondary piece?
Michael Moeser (21:59):
Makes sense. Now you talked in 2024 about a future where there's a financial autopilot where borrowers are proactively guided into better financial products. How close are we to that reality today?
Nima Ghamsari (22:15):
Well, this kind of goes back to the example I was mentioning around every consumer's financial situation is different and a very robust, very experienced loan officer or banker could walk you through why this product or that product makes the most sense for you. But that has two major flaws, which is one, the consumer has to know to reach out in the first place for you to be able to help them. And two, you can't afford, let's say there's 10 million people in a year, in a great year, 10 million people get a mortgage. Who's going to spend all that time and energy with 10 million people in a high volume a year who can't afford to do it. And then the lowest income, the least valuable in quotes, consumer, because everything has variable economics, they get the least amount of time and attention, so they're not getting that in a scalable way.
(23:17):
But if you can make those humans that are serving that or make the systems better and make the humans superhumans and make the systems do the work in the background, that becomes really powerful. So that takes the reactive nature of a consumer reached out and wants help to, "Hey, I've done business with you before and here's a really great outcome for you given your current financial situation." So that makes the proactive possible and I think agents will help with that absolutely. Because if not, there are too many products, there are too many financial situations to be able to solve in the old-fashioned way that makes it reactive into proactive. And then on top of that, now these humans are super humans. And so as you learn more about the consumer situation in real time, the agent is working for you and helping you understand what might be an even better option than what was initially thought.
(24:05):
And so I think that means that anybody in the country, every single person in the country can have that high quality experience because now you have a hundred times the capacity in the industry.
Michael Moeser (24:18):
Makes sense. Now in a market where we don't have tons and tons of borrowers and applicants, every individual applicant is that much more valuable. How can lenders use AI and operational intelligence to better identify, prioritize and convert high intent borrowers? Any examples there?
Nima Ghamsari (24:42):
Yeah. And we haven't launched anything around this yet, but something that I'm excited about and we're working on is how do you help both the individual loan officer understand all maybe their history of customers and maybe their current customers
(24:59):
And that's a lot of information, again, to go back and look through and say, who should I reach out to today right now? What are the things that I'm working on? Well, how do I help them think about holistically, how do they help their broad maybe thousand people they've worked with in the past? What are the 30 that I'm working on now of those 30, which five should I reach out to right now because I need something from them to get this deal done for them? And for those other 970 that they're not working with right now, who got a mortgage in 2023 at 7% that I should be reaching out to with a very tailored offer? And so I think, or I should just even say offer, I'd say opportunity for savings for them.
(25:40):
And then I think for organizations, if you're an organization who's done work with millions or hundreds of thousands of consumers or you do work with them on an ongoing basis, so much of what consumers turn to organizations like that for is, how can you help me make my financial life better? You're a financial services institution, you're a bank or you're a credit union, you're your lender, how can you make my financial life better? And again, the old way of doing this of a rules-based approach to trying to figure out and putting people in broad segmented buckets of like, we think that a credit card makes the most sense for you or we think you're most likely to buy a car. We think you had a kid and you're most likely to go buy a home. That's the other beauty of this is that agents are so personalized and so that person's financial life and their specifics, that it becomes something where you can serve a customer with 100 times more care and personalization than you could before and as scale.
(26:41):
And it was hard to do at scale before and now it's easy to do at scale because of AI agents. And so something I'm super excited about and we're spending some time on.
Michael Moeser (26:48):
Now let's talk about, there's a contradiction here in terms of consumers, they want things more digital, they want it proactive, these personalized loan recommendations. On the flip side, and you mentioned it earlier yourself, there are times where people want human guidance, especially around major financial decisions. And for many people, that mortgage is the biggest financial decision of their lives. What does an AI-driven workflow with human review look like? How does this approach, I guess, how does it support borrower needs, compliance, and of course consistency? Well,
Nima Ghamsari (27:26):
I think the reality is that there are a lot of really great places that humans might want help and that lenders might want to intervene. There are really high value add places where there can be important things that either the consumer wants help with or the lender wants to help the consumer with. And then there are also things that are wrote work and that wrote work is something that I think is going to be the first bucket of work that becomes, and everyone becomes a superhuman in that. And where that rote work is being done for the underwriter, before the underwriter takes a look and they can do a pass that has a full file, an accurate file and a fully auditable file so they can see everything that happened and it's like, wow, I woke up to a really clean pipeline and there's these three things that I know I have to do.
(28:21):
I think the human's job shifts to that air traffic control that I mentioned earlier, review, oversight, judgment, not reading documents and staring and comparing and taking a third pass at something that has already been reviewed twice by two other people.
(28:43):
When that happens, these agents are going to make those underwriters' lives a lot better and they're going to make the consumer a lot happier because they're going to get a response faster. And when you're in a stressful financial situation as a consumer, maybe because your income went down because you had to switch jobs or because you moved to a new geography and you're trying to figure out how to make ends meet, that's when having fast answers and certainty as quickly as possible so important and that is not possible unfortunately without either today prior to agents was not possible without either a lot of human touch. So it was sort of reserved for the highest income people or unless you were in a very specific segment that was this fast lane. And so now I think it sort of opens the door up for everyone to get that level of treatment that they deserve.
(29:38):
And so I'm very excited for that world. I think it's going to make everyone's lives quite a bit easier. Nobody enjoys doing the stare and compare. Nobody enjoys waiting around for two weeks for their underwriting team to finally be able to get around to them because they have a backlog. There's more work by a large margin, even with the teams that are there today and even with the low volumes that are there today, let alone if volumes double, there's more Our work by a large margin than we even have a handle of. And so being able to take a lot of those rote things out of the picture or make them at least done in the background is going to be a huge, huge value add to the industry.
Michael Moeser (30:15):
Yeah, I can see from the consumer perspective, especially, I mean, I think at a base level, am I approved for the loan? Can I buy the house I've been dreaming about? I'm curious, do you think that AI will widen the gap between those top tier lenders that utilize it to get rid of the stare and compare the rote work versus everyone else? Will it widen that gap?
Nima Ghamsari (30:36):
Well, I view it quite differently than that actually. What I view it as, and this is where I said earlier on, speed is a moat. I think the gap between the companies that adopt agents doing background work ambient intelligence, the companies that adopt that, whether they're small or they're large, are going to be successful because they're going to be able to serve more people much faster and at a lower cost than they've been able to do historically. And that's all good things for the consumer too, because they can pass on some of those savings to consumer if they want. And then there's going to be the have- nots. I mean, there are still companies out there that are skeptical
Michael Moeser (31:20):
Of
Nima Ghamsari (31:20):
Agents or they're skeptical of AI. And it's not to say that everyone is skeptical. I think most CEOs of large financial institutions I talk to are all in on it, but there's people on their teams who are like, "Well, we got to pay attention to this thing." And like I said, there's hallucination and there's potentially some data. Do you want to have data security? You want to make sure data is really secure and you want to make ... There's a lot of things to think through, but the organizations that are able to get through those really important things to get through but quickly aren't going to just do better than the organizations that don't. It's unfortunate because now imagine two companies, one with an unlimited elastic workforce and one without it. Which would you assume would do better? And so that's where I'm trying to help.
(32:09):
I think it's so important for us to help the industry. It's actually nice for me because it's a win for everyone. It makes the individuals working at these organizations superpower, superpowered, and it also makes the consumer's life a lot better and it helps the economics of the transaction. I mean, it makes everyone's life a lot better. And so I don't feel like there's a trade-off here. I know AI sort of gets a bad rap. It's like, "Oh, it's replacing human jobs." Actually, I think it's going to create more human jobs than ever in history because no company has ever said to their board, "Oh, we're done with our roadmap. We're done with everything we wanted to do. " We've been dreaming of this moment. We've always been short-staffed because it was expensive. I would love to have thousands of people working at Blend. I can't do it.
(32:54):
The financials don't work out for us. But now it's something that I can have that. I can have that and that is so important. And that just means so many new opportunities are going to get created. So many new products are going to get created, so many new services are going to get created and all of those things are going to need human oversight and they're going to need somebody to be able to step in when the buyer of that service wants to talk to a human, relationship management aspects. And so I think it's going to create more jobs than we've ever seen in history. I think the 2030s are going to be a decade of boom for the economy, unlike anything we've ever seen before because organizations are going to accelerate into the late 2020s and there's going to be so many new things that are created that we on this call can't even begin to fathom.
(33:40):
And that is a really exciting time.
Michael Moeser (33:43):
So as we wrap up today's conversation, Nima, are there any closing thoughts you'd like to share with our listeners?
Nima Ghamsari (33:50):
Yeah, I mean, I feel like I've been promising this possibility to our customers, the banks and lenders that are out there for 15 years almost. We started Blend in 2012 and we wanted to make it so that a consumer could get through the mortgage process again with the lender's oversight and be able to get through it fast and easy and with very little manual effort, manual labor on either side. I feel that now with Agentic AI, it's not just a vision. I mean, we had made such great progress, but it felt like we'd hit a ceiling because like I said, there's thousands of pages of guidelines. There's thousands of different financial situations. There's so many systems you have to touch and now Agentic AI can help solve that last mile problem, which is so important. And so I'm so excited about that. And for lenders who are listening on this or banks that are listening to this, I do think that what I said earlier around it's not about size, it's about can you adopt the tools that are being given to you fast?
(35:04):
And I don't even mean just Blend. I mean, there's probably all parts of your business that are thinking about Agentic transformation. It's how I think about Blend. We have software engineers, we have accountants, we have salespeople, we have customer service reps. I think about every role in our company and I'm thinking, how do I get every single one of those people to have an army of AI agents working for them? And I think that'd be my encouragement to the people and as quickly as possible, because if I'm able to get that to them today or tomorrow or this month or next month, that's a big deal. But if it takes me two, three years to get it to all of them, I've sort of fallen behind my competition, which is other technology companies. And so that's the same thing I would encourage our customers and partners and prospects to think of as they're thinking about their journey around banking.
Michael Moeser (35:50):
Super. Well, that's all the time we have for today. Thank you, Nima, for joining us again to share your perspective on AI's impact in the mortgage industry and thank you to our listeners for tuning in. If you'd like to learn more about Blend's autopilot program, please visit blend.com. Thank you. Thank you.