Can AI Fix Housing and Healthcare Affordability?

We're about 5 million housing units short of what we actually need in the country.

Our goal is to enable fully autonomous buildings. An entire portfolio has the ability to run core operations without requiring human intervention at all.

Housing and healthcare are the biggest expenses that people have.

It's pretty self-evident that technology makes the experience better for everyone and brings down costs, and more people should be working on it. Alex, we're honored to announce our latest investment in EliseAI. Minna, Tony, welcome to the podcast. Welcome to the portfolio.

Why don't you give the audience a brief background on why EliseAI, why housing, and why healthcare?

We wanted to use AI to solve real-world problems. Housing and healthcare are the biggest expenses that people have. They eat up about 42% of what a typical household makes, and nationally, these sectors make up about 40% of the entire GDP.

So it's pretty obvious that we need to figure out how to cut waste in these industries. It was pretty crazy to us that not enough people were working on these problems. Top technology people, for example, are not asking the hard questions about how we can actually fix them. It's just an accepted tax that we all pay.

So we started digging a lot deeper into how these systems operate, and once you do that, you can see inefficiencies everywhere you look. Those really add up to huge costs for all of us.

It's also not just about the money. The quality of these services has a really big impact on people's lives, and when they fail, it doesn't just hurt individuals; it actually hurts society as a whole. It's a fundamentally important problem for us to solve.

Alex, why don't you give the perspective from the growth team? There have been a lot of players in the space over time. What got you excited about this space, and why this team?

As you think about the space, a lot of it ties to what Minna just said. Our partner Mark, every year like clockwork, sends out this updated chart, and it shows consumer goods and services over time. If you look at housing and healthcare, as Minna just said, prices just go up and to the right. Any industry that technology has touched has gone down.

You can see that in computers, TVs, and video games. We keep asking ourselves, why hasn't software eaten this industry? Why hasn't software touched housing and healthcare?

Right now, it's such a paradigm shift with AI, where the operational, administrative, and communication burden can really change. You look at Minna and Tony: 2 incredibly tenacious technical co-founders. They were early to this problem. They started in 2017, and they have gone deep, first in housing and more recently in healthcare.

The results speak for themselves. Customers are absolutely delighted with the product. You see that qualitatively in the feedback we heard, but then quantitatively in the scale, the growth, and the efficiency. We're thrilled to partner with them.

Minna, why don't you go deeper in terms of what the unlocks are to actually improve housing affordability? Why don't you give us some more context on the problem?

For housing affordability, housing supply matters most of all. That's the single greatest determinant of housing prices. We know that we need to build way more units. We're about 5 million housing units short of what we actually need in the country, and we need to add somewhere between 1.8 million and 2 million units per year just to keep that shortage from getting worse, let alone making up for that deficit.

We've only developed about 1.5 million units last year, so we actually need to increase our delivery by about 50%. Analysts say that the pipeline is shrinking for 2026 and beyond. They said it's going to drop by about 50%, so we're headed in the completely wrong direction.

Housing supply matters, but in the short term, we can actually get more out of our current supply. I'll give you an example: almost half of the inquiries that go to a rental apartment building never get a response.

We've all experienced this, right? We send a message because we want to look at an apartment, but no one responds to us. We get ghosted, and that apartment is still sitting there available. You want it, but it's being underutilized just because the process is broken.

If AI is managing all that demand, we can turn vacant apartments into occupied apartments much faster. We have data that proves this works. Earlier this year, we provided data to an organization called ALN and found that buildings using EliseAI had 2% higher occupancy compared with the market.

We're working on fixing the affordability problem and the supply problem from a bunch of different angles.

You mentioned that we're going in the wrong direction. Can you share more about why that's happening, and what are the biggest regulatory or technological bottlenecks that need to be addressed for that to change?

There's a lot of regulation and a lot of zoning laws that cause part of this problem. But even relaxing regulation will not solve that problem by itself. We also need more capital flowing into housing construction today, and that's a big thing that we think about internally at EliseAI.

Capital flows where there are the highest returns. Right now, housing has lower returns relative to other asset classes. But we're actually enabling housing to achieve higher returns because we're creating 10-times-better housing operators that return higher profits to investors.

If housing is a stable, higher-return investment, how much more capital would be flowing to this most fundamental need for society? That's really great for everyone because it drives more supply into the market, and supply is really king when it comes to fixing the housing crisis.

On the regulatory side, do you think we'll see full YIMBYism in our lifetime? Or is your view, “Hey, there's some structural reason that impedes that in the U.S. relative to a country like Japan, and we just have to work with what we've got”?

I think we really hope so. Different cities have started making real progress toward this. Minneapolis is a great example. They did a big housing reform, and part of that was ending single-family zoning rules there. That happened back in 2019, and since then, we've seen supply grow 3 times faster than the national average. Rents have stayed flat for that whole period, compared with everybody else, who experienced about a 31% increase.

I think we're seeing some early signs there, and hopefully other cities and states take that example.

Even if we got Tokyo-level zoning tomorrow, how fast could supply respond? Or is it really just the capital balance that's preventing that?

I think the data is somewhat clear that it will obviously take a few years to see the full impact. But if you let people build, they will go and build, and the market will take care of itself. Obviously, there will be a lot more innovation as it becomes a much more competitive market.

Why don't we go deeper into understanding what needs to be true for housing to be a more attractive asset class? Maybe you can walk us through how it's made up and what would need to be true to improve the rate of return there.

It's very inefficient. That's the reality: we're dealing with the physical world, and along the way, real estate has not invested a ton in technology. So it hasn't gotten a lot of the efficiencies that an internet company or a SaaS company could achieve.

There are a bunch of headwinds working against this asset class. One is labor. Labor is super expensive, and it's only getting worse, particularly after COVID. There are a bunch of other reasons, like insurance premiums, costs, and supply-chain disruptions after COVID as well.

All in all, I think the biggest controllable expense that our customers look at is labor. There's not too much they can always do about it; it's not as low-hanging fruit as insurance changes.

New York and San Francisco are 2 markets that have really struggled with housing supply. Assuming we don't get more housing supply in these 2 cities, what do you think can be done to increase affordability?

Even if we don't build more units, San Francisco's vacancy rate is about 3.5% today, so there's definitely room for more efficiency. You can increase that utilization by making units turn over faster and by filling units faster. For all of that, better technology can help. You want to cut all the manual inefficiencies.

There are also other ideas: you can do smaller apartments, share amenities, and have more flexible layouts. Better infrastructure in general also helps. If you connect Jersey better to New York, obviously that increases the supply in the whole metro area, and that increases affordability.

I think all of these things can help a lot. To a certain extent, those are band-aids. We do need to build more units, and that will be the main way to consistently drive affordability. But I would say there's definitely more room for improvement even at today's supply.

Yeah, of course, we hope that more supply gets built and more red tape gets cut in San Francisco. I know Erik and I are at least optimistic about PermitSF, led by Mayor Lurie, but absent that, software probably can be an important lever for resolving some of these issues. Where do you see that tangibly impacting affordability?

I think technology can counter some of the cost inflation and headwinds, but it's also a sector that historically hasn't invested much in tech compared to other industries. So, number 1, we have to execute really well to get these sorts of non-tech-adopting audiences to use what we build and accept it. But if we do that well, the biggest controllable expense is labor. AI being used to automate and eliminate a lot of the manual and inefficient workflows can help a lot.

We're already seeing customers reduce some of these expenses and find other savings as well, like cutting legal fees because their operations are already more compliant from the start, or reducing CapEx costs by optimizing preventative maintenance so things don't break as much. All of these things add up, and if someone's absorbing those costs and it's you, then, yeah, we're creating these sorts of 10x operators.

My landlord is not thanking me. We've seen many of your clients be able to centralize a lot of their staff, increase AI use as a communication mechanism with their tenants, and the impact has been dramatic. I know with Equity Residential, one of your customers, they've gotten up to 200 units per employee. Maybe just walk us through: How do you get from half of that, which I think is the baseline expectation, to twice as high? And then how do you think about the efficient frontier? Where should that be 5 years from now?

Our goal is to enable fully autonomous buildings. That means an entire portfolio has the ability to run core operations without requiring human intervention at all. When you look at what actually happens on-site at a building, you quickly realize how much of that day-to-day work is just administrative in nature and can be automated away. When you're thinking about the physical limits, it's truly the physical work that's left—the maintenance or things that are maybe legally required. Those are the parts that I think are a little bit harder and actually put the thresholds on what's possible today.

Going after full automation is a really hard technical challenge that no one has truly figured out yet. We really have to reimagine our customers' entire operating models to be able to develop products that support this. Equity Residential was a great example. Really early on, they had taken advantage of a ton of technology to get cost optimization and efficiencies.

Brookfield Properties is another one of our customers. They're building this centralized model that enables a single employee to service multiple properties, and they're actually finding that they can get a single employee to work across 10,000 units using AI in a specialized role. So think about that: 1 person managing what used to require dozens of people who were decentralized across multiple properties. It's a big difference, but you need a huge amount of automation to be able to achieve those numbers.

Even the boundaries of what's actually physical are changing fast. Something like door access wasn't a thing 10 years ago, and today we see more and more physical keys being replaced by smart locks. Now, you connect your AI to that system, and you can do key provisioning online. I think we'll see more sensors in the buildings, more decisions being driven, and more planning being done by AI. Of course, there are still some physical boundaries. The AI isn't going to fix the sink, but I think we can push that boundary quite a lot in the next few years.

Why don't you talk through where we are today in terms of what's been automated and what's not yet automated? Then we can get to the full vision. In that full vision, what do humans do?

Yeah, a lot has been automated. Maintenance used to involve physical boards covered in Post-it notes, with people trying to keep track of what needs to be done. Now that can be automated, triaged by AI, prioritized based on the urgency of those issues, routed to the right technician, and tracked automatically. We see that some operators with these new workflows have cut average work-order completion times from 4 to 5 days down to under 48 hours. That's really meaningful for residents.

Leasing was maybe one of the worst. People were spending entire days answering emails—the same 50 questions over and over all day long, just the same basic information. Now AI can obviously handle those tasks with all the knowledge of a building or all the knowledge of a portfolio.

Touring was another area of automation. You'd have to go meet a broker or a leasing agent and be physically escorted to every single showing, but now AI can give you access, or you can get access through smart hardware, smart locks, and lockboxes. The AI can still be there to engage, answer all those questions, and do the selling. That gives you a ton of benefits and efficiencies because you're not just paying a whole human just to unlock a single door. It actually cuts down the average time for our customers from about 30 days from listing an apartment to leasing it to under 14 days, because it gives you so much more flexibility to tour around the clock.

Documentation was a big area of automation, too. Lots of people were just copying and pasting fields. It was just an industry that was super ripe for technology and didn't get it for a very long time.

Yeah. And I think this is only even the first-order automation we can do. I think there's a whole second order, because what Minna is talking about today is on that single-community level, a single building. There's this whole question about how you organize the entire ecosystem: How can you share resources—whether it's people, parts, or tools—between many buildings? Obviously, that increases the complexity, but for something like maintenance, we expect to see a lot more dramatic gains once you go to the ecosystem level.

And how about the second part of the question? In terms of, okay, your vision comes true—fully autonomous or fully automated housing—right now, there are a lot of people doing a lot of those activities, or some of them. What do they go do?

I think as AI takes over a lot of the communication and logistics, human roles don't all disappear. These sorts of AI-enabled career paths start to emerge. And I think you're seeing this in other industries as well. Expectations are higher. People, like you mentioned earlier, expect that renting an apartment is super easy, super frictionless, and it doesn't require extensive human contact for basic stuff. That's a great opportunity for AI.

I think you'll see that these career paths—or, in the short term, the menial parts of the job—go away, and people spend time focusing on building relationships with residents and creating communities. I think people are spending more time at home, right? They're working at home; their home is their office. You still need this human connection, and a lot of our customers are creating roles like community engagement and helping people socialize.

I think people will have specialized tracks. Instead of being a generalist on-site doing everything menial and complex, you'll actually see a lot more specialization. You might become a renewal specialist who handles the trickiest retention cases or a resident-experience specialist who resolves conflicts. But I think in the long run, people will be managing big workforces of AI and overseeing these automated systems that are running most of the work.

Yeah. And I'll add to that: On the maintenance side, obviously, I think that physical aspect is never going to fully go away. I think it'll be a lot more efficient, which is really needed, because there's just huge shortages from a labor perspective, and a lot of the maintenance technicians today are actually over 50 years old. So you can see that in the next few years, it's only going to get worse and worse if there are no more efficiency gains.

Going back to just the future of housing for a second, let's say 10 years from now we have robotics, we've made some major advances in longevity research, and we have AGI. Why don't you paint what the experience is going to look like, or how those impacts—the ones that, at least to me, are robotics and longevity—are going to impact the multifamily asset class?

I think the population is going to change a lot with AI. The reason I talk about longevity is that I think a lot of AI research is—and should be—going toward massively extending human life. If you're living longer, there are more people around staying around longer.

We talk about cost of living a lot. Cost of living is actually the number 1 reason that people don't have more children. So if AI creates a lot more wealth in the world and brings down the cost of living because of efficiencies similar to what we work on in housing and health care, people will have more kids. Again, you have more population.

So all these things, I think, will really affect the housing market because it is a fundamental need for all those people. We have to find efficiencies. As I mentioned before, that means more housing supply, and that's where robotics can come in and play a huge role. Can we use robots in manufacturing? Can we do modular housing? Can we build faster with robots? Can we bring down the cost of construction? I actually think that's one area that we've never touched. We're not a hardware company yet, but it is this big piece of the puzzle in the long-term vision that really needs to be solved by somebody.

I think you also want to leverage the technology to increase mobility in general. Today, people are stuck in these 12-month or 24-month leases. In the future, you want to have the flexibility to just move in tomorrow, and for that to be very cheap. You can have a much greater degree of flexibility.

Yeah, I think mobility in the market is incredibly important and productive for society. If you think about it, we are signing these 12-month leases and locked into these contracts as consumers. That's a big commitment. A lot of people don't want to sign these leases, but it's really laborious to turn over every apartment—not just the maintenance turnover, but finding somebody, answering all these same 50 questions over and over again, and touring. There's so much labor that it makes it really difficult for an operator, like a housing operator, to find a new resident or tenant.

AI doesn't really care if it's signing shorter-term leases. It's just doing that over and over, and it scales. Then you get the benefit for both the landlord and the consumer: they can do it at a cheaper cost, and people can be more mobile. I think that opens up a ton of opportunity. People can move for jobs more easily, move for their children's schools, and improve their quality of life. There are a bunch of different benefits to mobility. We've been talking about technology as a lever here and the need for it in real estate, and yet real estate spends the least on R&D of any other industry. Why is that, and what could be done about it?

I think solving housing operations is incredibly hard. The search space is massive; there's just a ton of different edge cases. We face this every day. Every single building is different, and it's a really large expense for people. All these things make interactions with an apartment, like leasing, really complex.

In the past, you really just needed a person because traditional software couldn't handle the variability that was required. If you needed a person anyway, there wasn't really a motivation to buy technology—you just leaned on that person. You leaned on people to do absolutely everything. Of course, those people got really overburdened, but in the meantime, a lot of that critical data was never collected because it just lived in people's heads.

Now AI has changed what's feasible. It can handle these really large search spaces and these really complex operations. In addition, real estate is so far behind on tech that it actually has the most to benefit from AI. It may be going from the lowest R&D spending to potentially one of the highest spenders on AI because AI has finally unlocked automation that was possible but that they couldn't take advantage of or optimize before.

How do you respond to critics of proptech who say proptech is here to help residential real estate companies extract more value from their tenants?

That's honestly a pretty silly argument. I get it. Housing is a really emotionally charged subject. People apply these irrational expectations to landlords that they never apply to other types of business owners.

Think about other industries: you wouldn't want airlines to stick to paper ticketing processes so that they don't extract more value from tech efficiencies. You wouldn't want a supermarket to avoid scanners or barcodes. It's pretty self-evident that technology makes the experience better for everyone and brings down costs. We actually want landlords to use as much technology as possible because when they're slow to innovate, that's actually bad for consumers.

The barriers to entry are already quite high. The operations are very complex. Everything you have to do is multimodal, and so much of it is manual. I think that limits the number of people who can get into that business in the first place. I think that actually gives a lot more pricing power to the existing landlords.

You can make the argument that all of this inefficiency is really bad for the consumer. At the end of the day, we believe that competitive markets take care of themselves, and I cannot think of a single example where technology was banned and then costs went down. I think that just never happens.

Yeah, it's exactly the opposite, right? Generally, when technology is introduced, you see a surplus, and most of that typically goes back to the consumer. The criticism of proptech would be: is it all going to the property managers and the owner-operators? Hopefully, as more and more AI is adopted, it's going to address this affordability crisis.

You need mass adoption. That's where the competitive markets take care of themselves. There are rising repair and maintenance costs. These keep vacancies longer. Apartments aren't occupied, and this leads to higher housing costs as well. What is EliseAI doing to address this? What can be done over the next 5 years?

Obviously, there's a very physical component that needs to be taken care of. But where we think AI and technology more broadly can help is that all of these problems are actually very complex planning problems that are quite difficult to solve with the current level of technology.

There's so much computation that you need to do around how to get smarter at scheduling technicians and routing, and how to embed a lot of these common-sense things that everybody on the ground knows. If you're fixing a dishwasher, somebody can go and fix the holes in the wall at the same time. But if you actually want to paint the wall, first you need to fix the holes; you cannot just paint over them.

There's a lot of technology that we believe we can build that will make all of these planning, purchasing, scheduling, and orchestration decisions so much more efficient. We're not even touching what we think can be another big needle-mover around the preventative aspect: how do you track what's going on in a property? How do you know when appliances are nearing their end of life and replace them in smart and cheap ways? We feel like all these problems can have a really big impact, because today we see our clients waste days and days because some piece of information got lost.

Every day you shave off the average unit-turn time nationwide unlocks billions of dollars in value. There's a ton of value just by moving the needle a little bit. The causes of those delays are completely avoidable. This part wasn't delivered, or the data wasn't input into the system, so the next person wasn't scheduled for that job. It's totally addressable through automation, and those are the problems that I think are really exciting.

When I met Minna in 2021, she and Tony were building a relatively narrow tool for leasing. We were catching up a bit over a year ago, in 2024. Leasing had expanded to broader residential operations—maintenance, which we've talked a lot about, billing, delinquencies, and so on. Then she dropped, “We're launching in healthcare.” We launched in healthcare, and I thought to myself, “Minna, Tony, that's insane.” What do those 2 industries have in common? I dismissed it, but kept it in the back of my head.

Then, catching up a couple of months ago, the healthcare business was humming. Maybe share what is similar in the workflows, what is different, and what's been most exciting there.

I agree with you that those 2 fields look very different. We've been mostly touching the admin piece of healthcare, and we found those problem sets are quite similar. You see these very bloated cost structures that have so many inefficiencies. They're all struggling with staffing, and all of these total costs get pushed to the end consumer.

We think a lot of the causes of that are the very similar dynamics we see between these complex digital-physical interactions that are full of regulations. We believe AI can help quite a lot in both of them. We see so much commonality around how they approach things like intake. You have to collect very structured information around names, preferences, budgets, and insurance. You keep dealing with this really high volume of repetitive inquiries.

You get the same questions again and again. It's all done over the phone, through conversations. We've been able to adopt quite a lot of our technology pretty seamlessly. We developed our voice technology in the housing space, and that has translated really well to healthcare. The same thing with a lot of the scheduling optimizations we've been doing has translated quite well. They feel very different, but from an administrative or operations perspective, we've barely been surprised by anything in transitioning to healthcare.

We spent a bunch of time earlier talking about why housing costs are so high. It seems like, no matter what, healthcare costs remain at 1/6 or 1/5 of the economy. Is it that we're just getting a better product for that cost, and because healthcare is a good that we just keep wanting more of it? No matter what, we get something better and we're just going to keep spending. Is it inelastic that way, or is it that some morass of regulatory challenges prevents technology from bending that price curve? What's happening there? Why are costs staying the same?

I think both are actually true. Healthcare is definitely a very elastic need: people are getting better healthcare, and people are getting more healthcare. I think if costs go down, people would want even more healthcare. I definitely think that's a good thing, right? It makes people live longer, happier lives. That's super important, and that's all true.

On the administrative side, I don't think we're getting a better administrative experience. I think the costs on the administrative side have really skyrocketed way faster than anything on the clinical side. Partially, people have invested in technology, but it just hasn't been quite good enough, again because so much is happening over the phone in these unstructured interactions. We think that, with the current level of technology, you can actually make a really big dent in the administrative aspect.

I don't think there's been this huge boom of technology adoption in healthcare that we're surprised hasn't flowed to consumers' pockets yet. I think we'll see that with AI, but I don't think the cycle has given its feedback yet. I'm still hopeful. It can arise in either lower costs, better outcomes, or a combination of both of those things.

In housing, you guys have really gone from leasing to broader resident operations. You've started with scheduling on the healthcare side. Where do you think the platform goes from here?

We think healthcare is much earlier for us. There's so much more happening on the administrative backend side of things that we feel like there are so many inefficiencies that need to be addressed, and I think that's going to take a bit to cover all of that ground.

I think it could be anything from that first interaction to the billing cycle and, more importantly, how you keep the communication going post-appointment with the patient. Today, you go in and spend 10 minutes with the doctor. They're definitely very helpful, but then you get a piece of paper, and it's good luck from there onward. AI can help quite a lot with engagement on the patient side, and I think there's a lot to be done there.

You go home from your appointment, and you have 4 things that you're supposed to do every night for the next week. What is adherence to that? Pretty low. But if you got an EliseAI message every night, you'd probably do a better job.

I think AI plays a great role in education and leaving the outcome with the patient. If AI can scale and achieve better treatment plan fulfillment, that's going to be better for everybody. It's certainly going to save us a lot of costs. It's one of the government's largest expenses, and so we're all paying for those outcomes being poor as well.

This adherence is not easy at all because, again, patients are very stressed when they go to a doctor. It's not an easy thing to do, and I think AI can help by giving them more time after the appointment to ask questions. Obviously, there are things like language barriers that AI can help quite a lot with.

Yeah, and involve family members who probably weren't able to attend the appointment or the procedure.

You have to think about all your questions right in the moment when you have that time with the doctor, and then you think of something too late.

In this episode, we've been talking about housing and healthcare. These are 2 extremely complex markets. I'm curious: if you can go back in time, knowing what you know now, what might you have done differently?

I probably would have started with affordable housing, actually. Affordable housing has kind of every problem that the rest of the industry has, plus maximal complexity because of all the dense compliance and paperwork and all the additional requirements. They are the most underserved. They have the biggest administrative drag, and they are also some of the slowest adopters.

It just shows that there's this huge, clear opportunity for them to take advantage of AI, and it just takes longer to get there. That's one big thing I would change. We're actually approaching healthcare in a similar way, which is to start with what is the most underserved, because that's where we can have the largest impact: the most underserved and the most complex. If you solve those problems, then the rest of it is easier downstream.

Perhaps let's close on the ultimate vision for EliseAI. If you achieve everything you're setting out to do—and obviously, you've achieved a ton to date—what more can you say about what that looks like at scale?

I think our drive always has been cost reduction. If, at some point—and obviously, it's not just 1 company's effort—we get to a place where housing and healthcare are not cost concerns for the average person, I think that would be amazing.

If we can take this 42% of what a household spends on housing and healthcare and bring that down to 20-some percent, that is, I think, one of the largest, most important problems we can solve, and more people should be working on it.

That's a great note to wrap on. Minna and Tony, thanks so much for coming on the podcast and being part of the portfolio.