- 7 weeks ago
On today’s episode, Editor in Chief Sarah Wheeler talks with Nate Baker, CEO of Qualia and the founder of 150 tech companies, to talk about the state of AI and what companies should be doing right now.
To learn more about Trust & Will, click here.
Related to this episode:
Artificial Intelligence
https://www.housingwire.com/tag/artificial-intelligence/
HousingWire | YouTube
https://www.youtube.com/channel/UCXDD_3y3LvU60vac7eki-6Q
More info about HousingWire
https://lnk.bio/housingwire
The HousingWire Daily podcast brings the full picture of the most compelling stories in the housing market reported across HousingWire. Each morning, listen to editor in chief Sarah Wheeler talk to leading industry voices and get a deeper look behind the scenes of the top mortgage and real estate stories. Hosted and produced by the HousingWire Content Studio.
To learn more about Trust & Will, click here.
Related to this episode:
Artificial Intelligence
https://www.housingwire.com/tag/artificial-intelligence/
HousingWire | YouTube
https://www.youtube.com/channel/UCXDD_3y3LvU60vac7eki-6Q
More info about HousingWire
https://lnk.bio/housingwire
The HousingWire Daily podcast brings the full picture of the most compelling stories in the housing market reported across HousingWire. Each morning, listen to editor in chief Sarah Wheeler talk to leading industry voices and get a deeper look behind the scenes of the top mortgage and real estate stories. Hosted and produced by the HousingWire Content Studio.
Category
🗞
NewsTranscript
00:00Welcome, everyone. My guest today is Nate Baker, CEO of Qualia, but also a serial tech entrepreneur
00:12to talk about the state of AI right now and why we really are at a tipping point when it comes
00:18to this technology. First, I want to thank our sponsor, Trust in Will, for making this episode
00:23possible. Nate, welcome back to the podcast. Thank you for having me. Excited to be here.
00:28I'm so glad you're here. It's been a while. And one of the things that is so unique about you is,
00:34yes, you run Qualia, but you have a whole lot of other projects going on at the same time
00:40from a tech standpoint. Would love to know some of the things that you're doing there.
00:45Qualia is an example of vertical SaaS, so software for a particular type of end market. Most people
00:53don't think about title agents very much. And so we came and built a software platform
00:5810 years ago for that category. I started Fractal, a different company, about six years ago to basically
01:06take that playbook to a bunch of other industries. We've actually launched about 150 companies
01:13that are in the style of Qualia in all of these other businesses like scrapyard software or rehab
01:21clinic software. And they're at various stages of scaling up. That is crazy. Okay. So your background
01:28is obviously tech and not necessarily title. Yes. Yeah. I learned about title by living in our first
01:34customer's basement for a while, Barry Feingold. But before that, I worked at a venture fund in San
01:40Francisco. And my boss at the time had started Palantir, a big enterprise software company.
01:45And that had gotten me interested in, you know, what I had thought of at the time as more obscure
01:52businesses. But when working there, basically went and looked for what are industries where
01:59there hasn't been sufficient technology investment. And that's how I found title initially.
02:06Yes. Well, we're glad you're here. Obviously, there's lots that has happened in the last 10
02:10years with Qualia and with the industry. I want to dig in a little bit on your larger tech background,
02:16because I feel like we are at such a tipping point. So, you know, we had our we started our AI
02:22summit two years ago. Well, last year, we just had our second one. And the difference in what we were
02:29covering in how fast things had changed in the maturity of people's understanding about AI in that
02:35one year was pretty incredible for me. So you starting 150 companies in tech would love to get
02:42your take on where we are relative to where we were.
02:47The the difference is two years ago, basically, ChatGPT came out and you could pretty much write
02:53poems to one another. And that was cool and fun. And about a year ago, it started to be clear that
03:01this stuff was going to actually work in a business context. I think one of the more interesting
03:05companies that was growing during that year is called Cursor, which is a developer tool.
03:10They grew in that year from zero to five hundred million of revenue selling software to developers
03:17who are almost definitionally early adopters of technology. But the big breakthrough for me
03:23was Cloud Code launched at the beginning of this year and completely changed what people thought was
03:31possible that an AI model could do and actually help you do your work. And they showed us a breakthrough
03:40user interface as well, which is more of like a chat interface like ChatGPT, but where it actually
03:47can do things and step make a plan and step through a bunch of tasks and then say, this is what I want
03:55to do. Is that OK? And get your approval. And for me, that was just an aha moment. And I think it's
04:03the same for a lot of other entrepreneurs. And so what we're seeing across technology right now
04:11is the biggest investment ever in new companies. And the difference is that in the last six months,
04:22most business use cases are actually getting good enough with AI, where six months ago they were not.
04:30Very concretely, we can see that with title exam. Six months ago, you could not automate title exam.
04:37It was not possible to do it if it wasn't like a rules-based system. Today, it is possible.
04:45And it's not that hard. And so I think that what's going to happen is just a tremendous amount of
04:50change. I was at a conference this week where somebody stood up on stage and said, the whole
04:57idea of AI isn't going to take your job, but people who use AI are going to take your job. And he called
05:03BS on that. And he was like, listen, there are a lot of jobs that are not going to be around
05:08because of AI. And at our AI summit, we had a speaker that talked about here are the jobs most
05:15likely in mortgage and real estate to go away, right? And hopefully people can find more interesting
05:21jobs to do. What do you think about that? It's going to be a mixture. I think a good example
05:27internally at Qualia is something that people have been saying is going to get automated is
05:33support, your software support. And we have automated a tremendous amount of our support
05:39internally. However, that doesn't actually mean we have that many fewer people doing support.
05:45We've just significantly expanded what is in scope that our support team ought to do.
05:51And so I think that to continue winning customer business in 12 months, you're going to have to do
05:57100x more for that customer to continue earning that business. That changes everything. And so
06:08maybe the way to think about it is not that that job is going to be fully automated. It's that
06:12what that job used to be is going to be fully automated, but the scope of what it needs to be in
06:18the future is going to be much bigger. That's, that's interesting. I mean, we've definitely I
06:23guess what I've heard from different tech leaders is that, you know, they can scale from this so that
06:29they don't necessarily have to cut, but they can scale. But it does feel like there are just whole
06:34whole types of jobs. And I'm going to now I'm going to quote you from your LinkedIn, a recent LinkedIn
06:38post. And you said in 24 months, it's hard to imagine that title companies or any professional in any
06:45field is routinely entering data, modifying documents or communicating on individual repetitive
06:51emails, which we all do a lot of business through emails at this point, right? And you were talking
06:57about going from a system of record to a system of action. We'd love to pick your brain a little bit
07:02on that and what you mean there. So this is going to happen, maybe even faster. I think that today,
07:12most of our users sit in Qualia and do data entry into Qualia, they take actions in Qualia,
07:18they send emails, and a lot of that is highly automated, you know, where they've set up workflows
07:23and triggers and all this. But at the end of the day, most of where our users are kind of getting a
07:32sense of what they need to do next is from their email, or a phone call or something like that.
07:39And fundamentally, if you have the entire context of what's going on in a real estate transaction,
07:45and you can give it to an AI system that wants to decide the actions to take on that email,
07:53it is possible. So what we launched recently with Qualia Clear is the path that this is going to
08:00happen through. If you pre-process an email and then say, do you want to save this date of the file,
08:06save this document to the file, send this reply, modify this thing, it goes from you doing all that
08:14stuff to the machine saying, do you want to do all this stuff, and you hitting yes. And that's not
08:22our innovation. I think that Cloud Code, to me, is really the innovator in that user interface.
08:29But I think that is how every business is going to work. And it's figuring out how to make
08:36the person, the expert, a reviewer and an editor, more so than the author of re-keying all this
08:46information. So right now, right? I mean, even if you're in Gmail, it'll be like, it'll suggest like,
08:53here's an answer or here's, here's a thing. So how is this different than that?
08:57That doesn't, the problem with that and the reason, I mean, I personally don't use it very much
09:02is that, you know, if I'm trying to reply to you and you've emailed me and we're setting this
09:09conversation up, I need to know, or the system needs to know, who are you? Everything about you,
09:16everything about my schedule, everything about my travel that might not be in my calendar,
09:22it needs to be able to synthesize all of that and then answer with the context of the entire,
09:30my entire world of what's going on. And if it's missing any of that, it will reply to you
09:37incorrectly because maybe I'm traveling and it schedules something that I can't be at. And then
09:44it's just really frustrating. So what's happened in the last really three months is that
09:51tool use, which is these AI systems being able to go out and ask your Gmail, ask your calendar,
09:58ask Qualia what's going on and getting a good enough answer back has improved sufficiently
10:05that you will be able to automate that reply accurately. And that changes everything.
10:13Yeah. The accurately is, is the most important part there. Right. Um, we all know that sometimes,
10:19uh, if a, if a system doesn't know something, it might make it up because it doesn't want to,
10:23you know, it doesn't want to disappoint you or maybe I'm not sure why it does that in every case,
10:28but I know you've talked specifically about it, Qualia, how you've built it so that it doesn't do
10:33that so that it actually flags you. And it's like, no, this is not right.
10:37Yeah. And it is currently inherent to LMs, you know, these AI models that they will hallucinate
10:45stuff. Um, and so you have to do your best to put controls around that and also push it to your
10:53user to say, is this correct? But it's very challenging to fully eliminate that today.
10:59That said, the error rate is declining significantly because of improvements the model providers have
11:07made. So one of the, one of the things that I felt like even two years ago when, uh, chat GPT was
11:13released, you know, there was a lot of that kind of like, should I jump in now? Um, you know, in the
11:19mortgage and real estate industry, people have spent a lot of money on technology already. Um, they
11:24maybe haven't always, you know, it hasn't necessarily lowered the cost of doing things or, um,
11:29especially in a low volume environment, right? Like you, you, you have a lot of things that
11:33you need, but like, you have to have a certain amount of volume to make those things make sense
11:37from a, from a profit standpoint. And, but what I thought then, and cause I interviewed tech
11:43leaders every week and what they said was like, if you're not a first adopter here, you're just
11:48going to be behind. How do you feel like looking at it now, people who jumped in six months ago,
11:54a year ago, three months ago versus waiting another six months?
11:59It is an, it's a very interesting question because a lot of what we built 12 months ago
12:04that took us months, then we could build today in days. Um, it's that much different. And so
12:12yes, maybe we would have saved ourselves some time by not building it back then. However,
12:20I think that you build a muscle of how to use these tools and how to incorporate them into your
12:27business. And I think pushing on that is absolutely essential because the next 12 months, the only
12:36thing anybody should be doing in their business is figuring out how to use AI to supercharge their
12:43business in all aspects of what they do. And we're thinking about that a lot internally right now.
12:48I'm sure you're thinking about it as well.
12:50We are. Yeah, no, actually, uh, we've hired a, a, a strategic content editor. Um, internally,
12:56we call her our AI editor to, um, use our internal GPT to query our large amounts of data that we have
13:04through all those data, housing, our data, real trends data, all sorts of things. Um, and, and get
13:10insights and get stories and is been incredible. I can't even tell you, like it, it blows my mind.
13:15And someone asked me at a dinner recently, well, what do you think she'll be doing in six months?
13:20This person we hired, I was like, I can't even tell you because things are changing so fast there.
13:24She might be doing something completely different, but amazingly valuable for the company at that point.
13:30That's correct. And I think that it is, it is the case that six months from now,
13:35what we're all going to be doing is so far beyond where we're at right now that it's,
13:41you know, it's just going to be a completely different conversation. And the one thing that
13:46you definitely do need to do to get there though, is get your internal systems in order.
13:51If you have a content management system, if you manage all of your, uh, you know, interview
13:56transcripts somewhere, if you, you need a clean system where all your data is organized,
14:02you want your email integrated, you want all this stuff, because
14:06the only question that's going to matter over the next two years is, can you directly hook
14:12AI up to run your business in an extremely, uh, concrete way where you take, you get to take
14:20advantage of model improvement. And if you can't do that, it's going to be really hard to stay
14:27competitive.
14:29That's really interesting. So let me ask you this too, because obviously one of the promises of technology
14:35over the last couple of years is like, maybe it could democratize things, right? Like make the
14:40little guy more competitive against the big guy. Do you think AI accelerates that or is, is,
14:47proves that wrong in your opinion?
14:49I think it's a bit of both. This is a period, the next two years is going to be the
14:54highest variance two years I've ever seen. I think that if you're on the ball, even if you're
15:00one or two people, you can do incredibly well. Uh, and if you are a big company and you're not
15:06paying attention, you go out of business in two years. And I think that is true for software
15:12companies like us. It's true for mortgage company or a title company, a real estate agent. Um, you
15:18need to figure out how to use these tools to actually run your business. I spoke with a, a small, uh,
15:25title agent in Arkansas today who built by himself some incredible AI tools for, uh, making a commitment
15:36automatically, not a software engineer, just an expert in the industry. And so that's, that's
15:43amazing to me. And if you can, there's going to be a lot more of that, uh, but also there's going to
15:50be companies like us that are established in the industry trying to push really fast
15:55as well. So it's just going to be, it's going to be an interesting and dynamic environment.
16:00Boy, that that's crazy. That's why I feel like really, I know that probably in other times
16:05of like, um, really big technology leaps forward coming online. It's like, oh, this is it. But
16:12like, it just feels very different. It, to me, it feels like we are at a really different
16:18spot right now. Totally. A concrete example, my, my business partner, um, he gave a presentation
16:26at the Vatican recently on, you know, AI and the change and a lot of senior people over there.
16:32And it's just interesting to me, like in prior technology cycles, I don't think the Vatican
16:38was at the cutting edge of jumping in and saying like, we need to understand where this is going.
16:44But the thing that's been so interesting to me since we launched our quality of clear product
16:50a month ago is everybody understands that they need to do something about AI. It's even, even the
16:58Vatican, but it's not clear, uh, to a lot of people like, what do I do? Uh, and so that's, that's the
17:06big question. And I think this is the year where everyone needs to figure out, okay, stuff's good
17:12enough to actually help me in my business. What do I do? And if you're out there and you're a
17:17business owner, especially if you're at a smaller company or maybe a medium sized company, but you're
17:20not like a giant company, how do you answer that question? How do you even know what's what, where
17:25to start? I think if you do not have a personal chat GPT pro account or like the, uh, any subscription
17:33account where you can turn thinking on or a Claude account where you can use Sonic 4.5, just sign up for
17:40them. They're not that expensive. It's alien intelligence from the sky. That's smarter than
17:46me and basically everything. And just use that and start getting it into your daily work and figure
17:54out how to do integrations into those tools so that you can call your email, your calendar documents,
18:00et cetera. And if you start doing that, I think it will become very clear, very quickly, the types of
18:07things you can do. I think that's such great advice. And actually I think that's what I see when I talk
18:13to people who are sort of at the cutting edge, that's where they started. They started with just
18:17their own. And then they were like, wait a second, I can do that. And then that there's all sorts of
18:21business case, you know, use cases that come to mind where it's like, okay, what if, what if all of
18:26these employees could do that? What if, what if it can do this? And just, you don't really know until
18:31you ask it until you see what it is. And until you push back a little bit and see how much better it can be.
18:36Totally. Because initially the feeling is, this is amazing. It can do everything. And then once
18:42you start using it enough, you realize the limitations and then you're kind of like,
18:46oh, I can't do anything. And then you figure out how to work around them. And then you start
18:50actually becoming really productive with it. Okay. So it's like the, the, the stages of grief
18:55only for AI, right? Or stages of relationship with your AI is like the honeymoon phase. And it's like,
19:01oh wait, I don't know. What about these things? And then you figure out what you can actually fix
19:06and what is not going to get fixed. And one last point I want to make sure I leave
19:12you with is the rate of change here is unlike anything ever. It's improving exponentially right
19:21now. The most important thing to look at, I really like this company Meter, M-E-T-R. They publish
19:28a benchmark that shows the duration of tasks an LLM can do, meaning how long does it take a person to
19:37do this thing? And then how confidently can an AI do that today? And a year ago it was, you know,
19:45minutes, like very, very short tasks. And now it's becoming multiple hours long that of a human task
19:52that it can just do out of the box. And that has changed significantly this month. Even when we
20:01launched, uh, quality clear a month ago, we're using a Gemini 2.5 pro, which state of the art at the
20:07time. And since then you've seen GPT five, which is, which is a better model. And then you saw
20:14Sonnet 4.5 last week, which is, I think even better. Um, and supposedly 3.0 is going to come
20:22out in a week, uh, from Google. And I'm sure they'll, they'll be back at the top. And so just
20:28in the last month, it's been so dynamic, but it's hard to understand the rate of improvement if you're
20:34not using them all the time. That's a really good point. Again, I feel like this is one of those
20:41hands-on things and also maybe something that a leader needs to do and not delegate to somebody
20:45else to be kind of like, Oh, this is an it thing, or this is my technology people. It's like, I, it
20:51feels like to really understand it, you have to do it yourself. A hundred percent. If an example in
20:58title is I expect that in 12 months, you'll be able to go to chat GPT and type in, give me a title
21:05commitment for address. And if the records are publicly available, it will just be able to do it
21:12and it'll do it with higher accuracy than a person does today. Wow. And it basically be free. And so
21:20that type of thing is coming very soon. Um, and so the question is, okay, if those types of things are
21:29going to happen, how do you adapt to stay ahead of that and utilize that? Well, how do you do that,
21:34Nate? Tell us all. I think in some ways it's, it's hard to know exactly what the, the right
21:41strategy is. And I think it's going to be very business specific. Uh, but basically at any given
21:48time, you know, the model capability is at a certain level, you know, it's, it's only so good
21:54and you can build scaffolding or like tools around it to bring it a little bit further. And that's where
22:01a lot of the value is going to be generated over the next few years for AI companies and for
22:07businesses, which is taking the state of the art tools and figuring out how to make them even more
22:13useful for their use case. So I think that's going to be the whole thing. So how do you incorporate
22:17these tools into your business, um, and, and use them really well because not everyone's going to do
22:22that. Okay, Nate. So out of 150 companies, which is incredible by the way, um, how many of those have to do
22:29with real estate mortgage title, something, uh, affiliated with the business?
22:35Maybe 10 or 15, something like that. One of the best ones recently is a self-storage facilities
22:42company called Cubby is doing quite well. Oh, interesting. Okay. And how does AI, uh, apply to
22:49that? The interesting thing is that really every business has a lot of paper, a lot of communication,
22:56you know, a lot of tenant qualification or some sort of manual back and forth process. And if it
23:04happens in email and in documents, then, and it were over the phone, AI is just going to completely
23:09change how it works this year. Wow. That is so cool. Thank you so much for, um, spending time with
23:15us today. I always love to check in with you. Um, you're definitely a thought leader in this space,
23:20really, uh, not just a, a thought leader, but an action leader. Maybe we should say just, uh,
23:24turning out all these, uh, different software companies and, and applications. So thank you
23:30so much, Nate. Thank you for having me. I really appreciate it.
Be the first to comment