Skip to playerSkip to main content
In this conversation with HousingWire’s Allison LaForgia, Blend Founder and Head Nima Ghamsari shared how the company is approaching AI with an ambitious goal. Rather than using the technology to make incremental improvements to existing mortgage workflows, Blend is considering what the lending process could look like if it were rebuilt around AI from the ground up.
“I think the biggest mistake people make when going into AI is trying to be too incremental,” Nima Ghamsari said. “Our approach is: What would this industry look like from the ground up?”
That means looking across the mortgage process for areas where AI can fundamentally change the way things are done. “All the work that we do as humans to underwrite files, process files, talk to consumers, serve consumers the right product, we are saying we think we can make that a lot better with AI,” he said. “AI is getting better every day.”

#AIMortgage #BlendAutopilot #MortgageTech

Category

🗞
News
Transcript
00:06I'm Allison LaForgia from Dallas, Texas, sitting with Nima Gamsari, the founder and head of Blend.
00:12Nima, thank you for joining me today.
00:14Thanks for having me.
00:15So, Nima, tell me about how Blend is approaching AI and mortgage and your role in shaping that vision.
00:25I think the biggest mistake people make when going in AI is trying to be too incremental.
00:31So, our approach is, what would this industry look like from the ground up?
00:36And thinking, you know, all the work that we do as humans to underwrite files, process files, talk to consumers,
00:41serve consumers the right product.
00:43We are saying we think we can make that a lot better with AI.
00:47And the beautiful thing for us is that I think as an industry is that AI is getting better every
00:53day.
00:54So, as capabilities of ours collectively gets better and the industry adopts it more and AI under the hood gets
01:00better, we have a lot of tailwinds going our way.
01:04Blend is moving so quickly.
01:06We spoke only a couple of months ago and Autopilot hadn't been launched yet.
01:12You were just launching Autopilot.
01:14Since launching Autopilot, what is the biggest lesson that you've learned from working with lenders?
01:21Well, Autopilot was a totally different approach to products for Blend.
01:27We had originally built Blend as something that you would use and you'd get our team to come on site
01:33and implement and do for you to get you up and running very quickly.
01:37But with Autopilot, we wanted to get it out there as quickly as possible.
01:41So, we made it self-server to turn on, really easy, like one click basically to turn it on.
01:45I think we talked about that in our last podcast.
01:48And we weren't sure if that was going to work.
01:51We thought maybe lenders don't want to do that or they would be scared with AI.
01:55But it's been the opposite.
01:57Almost all of our, we had 50, 60, 70 lenders turn it on in the weeks after the last time
02:02we spoke.
02:03And we've done 50,000 plus loans on Autopilot already.
02:08And, you know, it's really exciting to see so many of our lenders, some of the biggest ones, many of
02:13the biggest ones, and most of the smallest ones start to really lean into it.
02:19You mentioned a couple of things, one click.
02:22I've also heard you talk about customization.
02:25Why is customization critical for AI when it comes to mortgage technology?
02:31Well, every lender has their own idiosyncrasies in how they either serve customers or when they want to ask customers
02:39of certain things or what programs and products they offer.
02:42We have, you know, a large customer who does a lot of VA loans.
02:46And so, they really wanted to make the VA streamline refinance process amazing with AI.
02:51And so, we have out-of-the-box capabilities for that.
02:54And then they can layer on their own overlays.
02:56But it's so important because if the way that AI is serving your business doesn't match how your operations work
03:02behind the scenes, the customer is going to get a really disjointed experience.
03:06And so, we didn't want that.
03:07We made it actually also very easy, to your point, to customize by just writing in plain English the things
03:12that are different about your organization.
03:14And Autopilot picks it up as part of its process and says, hey, I'm not going to ask the customer
03:18for this letter of explanation right now because they just want a pre-approval.
03:21I'm going to do that later because that's what my lender's overlay said.
03:26It's clear that you've had many, many conversations with lenders about what's needed, what adoption looks like, what successful implementation
03:33looks like.
03:35How do you build the partnerships with lenders instead of just delivering a software technology?
03:43Partnerships with lenders, you know, when you're a software company like us, in a vertical that's very concentrated in lending
03:51and mortgage lending, you have to think customer first.
03:55And for us, the way we built partnership over the last almost 15 years, 14 years, has been really caring
04:02about their outcomes.
04:03If they really care about how fast they can serve consumers, we need to care about that.
04:08If they really care about how much it costs them to serve consumers, we need to care about that.
04:11If they really care about recruiting the best loan officers, we need to care about that.
04:15And so the fact that we do actually care about those things and then we show up and when there's
04:21issues, we show up even harder has been, I think, the thing that's led to great relationships for me and
04:27for the team and for the company overall for the last 10, 15 years, but also great outcomes for our
04:32customers.
04:33There's going to be bumps whenever you're looking at implementing a new process, whether it's something as big and potentially
04:40as impactful as AI or it's any other technology.
04:43So I think that that partnership aspect and how you approach that is invaluable.
04:47Yeah, and as part of my job with our customers and our internal team, when the internal team is working
04:53nights and weekends to get this thing going and then our customers are adopting something new and for every one
04:59person that's excited about that innovation, there's one person who's terrified of it because their job is compliance or their
05:05job is actually serving the customers.
05:07And so part of my job with customers is to also serve a little bit as therapists and help them
05:13understand that, you know, we're going to work these things out.
05:17It's going to be okay.
05:19You know, I have to make sure that they understand that we're going to power through it.
05:22And that's also what helps build that partnership you're asking about.
05:25We've seen that AI adoption has accelerated.
05:29I mean, when you look at the past year, six months, three months, several weeks, what lessons have been the
05:38most important as you look at Blend's intention at driving meaningful change?
05:46Whenever we're trying to do something new at Blend, we start very close to the customer.
05:52And so the biggest, I think the reason Autopilot has been so successful so far doing 50,000 loans and
05:58making real meaningful impact to our customers' operations is we're just so close to the customer.
06:04We have meetings with every Autopilot customer every week.
06:08We're talking to them.
06:10The team is understanding them.
06:11We have a team that's dedicated 100% to it.
06:14It's not a part-time job for those people.
06:16And so that's the way that we're approaching this, just like almost like another startup within what I still consider
06:23a startup in Blend and making it so that that team has access to all the right things and all
06:30the right customers so they can make the best product.
06:32And we're all open-minded because AI is changing around us.
06:35One of the nice things about AI is that it's allowing us to treat the, you know, one thing I
06:40said in our last earnings call is that the output of code from our team has three X per person.
06:47Since January.
06:49When you can do that, anything is fair game.
06:53Everything, nothing that we've done in the past matters.
06:55The sunk cost, everything can be the future.
06:57And so I'm very excited about that.
06:58And the lenders have been great partners in helping us build that vision.
07:02That is incredibly exciting.
07:05I mean, even hearing about how hyper-focused your team is as they approach new things is amazing.
07:10Now, saying that, we have to go back to a word that you mentioned earlier, compliance, for a second.
07:16How do you balance innovation with compliance, security, and most importantly, customer trust?
07:24Well, compliance is paramount in an industry like this.
07:28And there's a lot of guidelines.
07:30There's a lot of regulation.
07:32And the nice thing is when there's lots of rules, if you can delineate those things to AI, AI can
07:39actually follow those rules for you.
07:42So I think part of it is AI will actually, over time, be better at checking and double-checking our
07:48work as humans and even its own work as AI.
07:52But in the meantime, one of the nice things about the foundation we're building on, we're not building on a
07:56blank foundation.
07:57And Blend's system has had a compliance foundation around RESPA, around ECOA, around all the words, all the acronyms that
08:05we're all familiar with as an industry, which we have too many of those acronyms.
08:09Too many.
08:09Way too many of those acronyms.
08:12It's already built with those things in mind.
08:14So if the AI helps a loan officer pull credit, ECOA automatically kicks in.
08:19The AI doesn't have to know about it.
08:21We're built on a really strong foundation at Blend.
08:23And so not only do we think it's going to be a good starting point, which it has been, but
08:28it's going to get better over time as AI gets better.
08:30So that's already built into the core way that you operate.
08:33It's built into the core way that our system runs.
08:35So our system can't, when trade is triggered, we can't not start that three-day clock.
08:40When credit is pulled, we can't not start that ECOA trigger.
08:43And then there's a privacy aspect as well, where a lot of our lenders are concerned about how do you
08:50make sure that the consumer's financial information doesn't get used to train AI in the future?
08:56Because it shouldn't be.
08:57And so that just takes good architecture and good engineering.
09:00And thankfully, that's something that I think we're maybe the best in the world at.
09:04Fantastic.
09:05Now, a hard question.
09:07What excites you the most about the future of AI in mortgage?
09:12Well, we pulled some early stats for Autopilot and we're about 19% higher conversion for consumers who go through
09:20an Autopilot process because Autopilot is underwriting people in real time.
09:25So when I said reimagine the industry to start the session, what I meant was in any other product, when
09:32you're applying for a car loan or a credit card, you're getting approved in real time.
09:36If there are things that are needed of you, they're asked of you in real time, and then that leads
09:40to an approval.
09:41And mortgage is different because we have so many guidelines, so many documents to look through.
09:46And now, for the first time ever, we can approve that consumer in real time.
09:50And so when I think of a world where that 18.8% better conversion we're getting because consumers are
09:57getting the approval in real time and the cycle time reduction of already three days from just three months of
10:03Autopilot rollout, we're just getting started and there's already meaningful impact.
10:08And I can see forward, you know, in my head, imagining a world where in five years consumers are going
10:16through, they're getting invited by the loan officer or they're applying online and they're going through and they're getting fully
10:20approved, clear to close in the first five, 10 minutes of working with the lender.
10:25And for a consumer who's probably making the scariest, most important purchase of their life, wanting to know that they're
10:33approved or knowing that they're approved is so important, making that as easy as possible.
10:37I think it's going to open up accessibility to consumers, just awareness when it becomes easier, when you can ask
10:44ChatGPT, hey, help me get approved for a mortgage.
10:46It can appoint you to a lender who can approve you in real time.
10:51It's going to, I think it's going to dramatically change how consumers even engage with this industry in a very
10:56positive way.
10:57I can't wait to see what's next.
10:59Congratulations on a successful rollout of Autopilot.
11:02You've moved so quickly up and I can't wait to see what you build next.
11:07Yeah, thank you for having me.
Comments

Recommended