Can AI Fix Housing and Healthcare Affordability?
Summary
EliseAI’s thesis is that AI can attack the two largest household expenses by removing administrative waste from housing and healthcare. Together they consume about 42% of a typical household’s income and constitute roughly 40% of GDP; Minna Song calls their inefficiency “an accepted tax that we all pay.”
Housing affordability remains fundamentally a supply problem, with software serving as an immediate utilization lever rather than a substitute for construction. The US is about 5 million units short, needs 1.8–2 million additions annually, and developed only about 1.5 million last year; meanwhile, analysts said the pipeline for 2026 and beyond would fall roughly 50%.
AI can improve effective supply by filling and turning units faster. Almost half of rental inquiries receive no response, while buildings using EliseAI recorded 2% higher occupancy versus market in data supplied to ALN. Customers also cut listing-to-lease time from about 30 days to under 14 and maintenance completion from four or five days to under 48 hours.
The investor flywheel depends on creating higher-return housing operators that attract more construction capital. Labor is the largest controllable expense, and EliseAI argues that “10x better housing operators” can raise profits enough to make housing more competitive with other asset classes—ultimately directing capital toward new supply.
The operating ambition is not incremental productivity but “fully autonomous buildings.” Equity Residential got up to 200 units per employee, while a specialized Brookfield Properties employee can work across 10,000 units with AI; the remaining limit is increasingly physical work because “the AI is not going to fix the sink.”
Automation may replace menial workflows while creating more specialized human roles and cheaper, more flexible housing. The founders envision renewal specialists, resident-experience teams, and people supervising AI workforces; by lowering turnover labor, AI could also enable leases shorter than today’s 12- or 24-month commitments.
Healthcare is the second market because its administrative machinery resembles housing more than its clinical surface suggests. Voice technology, structured intake including insurance information, scheduling, and repetitive inquiries transfer readily, while billing and post-appointment AI are additional areas the platform could address. AI could improve adherence, answer delayed questions, cross language barriers, and involve family members.
Consumer benefit is not automatic: it requires widespread adoption and competitive pass-through. The founders reject the idea that inefficient landlords protect tenants—“I cannot think of a single example where technology was banned and then costs went down”—and Minna Song’s long-run aspiration is that, if EliseAI can reduce housing and healthcare from 42% of household spending to “20-some percent,” they would solve one of society’s most important problems.
Deep dive
1. Supply, not software, is the binding housing constraint
Song’s starting point is unusually broad: housing and healthcare absorb about 42% of typical household income and roughly 40% of GDP. Their failures damage both individual lives and society, yet their inefficiencies remain “an accepted tax that we all pay.”
The housing arithmetic is adverse: the country is about 5 million units short and needs 1.8–2 million new units annually merely to stop the gap widening. Only about 1.5 million units were developed last year, implying delivery must rise roughly 50%, even as analysts said the 2026-and-beyond pipeline would shrink by about half.
Existing inventory can still work harder. Almost half of rental inquiries go unanswered, leaving wanted apartments vacant; buildings using EliseAI showed 2% higher occupancy versus market in data provided to ALN. But the founders repeatedly qualify this lever: operational efficiency is a “band-aid,” because “supply is really king.”
Song’s evidence for zoning reform is Minneapolis: after ending single-family zoning rules in 2019, its supply, she says, grew three times faster than the national average while rents stayed flat, versus a roughly 31% increase elsewhere. Even Tokyo-level liberalization would take years, but her call is categorical: “If you let people build, they will go and build.”
2. Better operators could pull capital back into housing
Relaxing regulation is insufficient without capital. Housing currently offers lower returns than competing asset classes, so EliseAI’s mechanism is to create “10x better housing operators”: automation raises profits, higher returns attract investment, and additional investment funds the supply that ultimately improves affordability.
Real estate missed the efficiencies enjoyed by internet and SaaS companies while absorbing higher post-COVID labor costs, insurance premiums, and supply-chain disruptions. Labor is customers’ largest controllable expense; secondary savings include lower legal costs through more compliant operations and lower CapEx through preventative maintenance.
Even without new construction, San Francisco’s roughly 3.5% vacancy rate leaves room to turn and fill units faster, redesign smaller apartments, share amenities, and improve regional connections. Maintenance orchestration is especially valuable: every day removed from nationwide unit-turn time “unlocks billions of dollars,” while many delays still come from missing parts, data, or scheduling handoffs.
3. Autonomous buildings turn administrative work into software
Stoyanov defines the endpoint as a portfolio whose core operations run without human intervention. Most on-site work is administrative; the practical floor is physical maintenance and legally mandated activity, although smart locks, digital key provisioning, and additional sensors keep moving that boundary. He calls full automation a hard technical challenge that no one has truly figured out yet.
Song describes maintenance as once involving physical boards covered in Post-it notes. AI can now triage urgency, route technicians, and track execution, helping some operators reduce completion times from four or five days to under 48 hours. Leasing automation similarly answers the same “50 questions” repeatedly and enables around-the-clock self-tours.
The operating benchmarks show the leverage: Equity Residential got up to 200 units per employee, about twice the cited baseline, while Brookfield Properties can have one specialized employee work across 10,000 units. That replaces decentralized coordination formerly requiring dozens of people, but only with “a huge amount of automation.”
Immerman calls today’s single-community workflows only “first-order” automation. The next gain, he says, comes from optimizing people, parts, and tools across many properties. Stoyanov illustrates the dependency-aware planning problem: holes must be repaired before a wall is painted, while AI could also track appliances nearing end of life and replace them in smart, cheap ways.
4. Consumer gains depend on competition, while jobs specialize
Stoyanov does not claim every job disappears. Administrative and logistics work contracts, while people specialize in difficult renewals, resident conflicts, community engagement, or oversight of large AI workforces. Torenberg adds that physical maintenance remains necessary—and efficiency is urgent because many maintenance technicians are over 50 and the labor shortage may worsen.
Lower operating friction could also enable shorter leases: as Torenberg puts it, AI does not care whether it repeats leasing work every month rather than annually. That could let people move more cheaply for jobs, schools, or quality of life; longer term, Song says robotics and modular manufacturing might reduce construction costs, although EliseAI is “not a hardware company yet.” Song also expects longevity and lower living costs to increase population and housing demand, making construction efficiencies more important.
The host’s proptech challenge—does automation merely help landlords extract more value?—draws a firm rebuttal. Stoyanov argues that inefficient operations raise barriers to entry and reinforce incumbent pricing power; Torenberg says technology’s consumer benefit depends on broad adoption and competitive pass-through. “You need mass adoption.”
5. The same administrative engine transfers into healthcare
Immerman recalls initially dismissing the healthcare expansion as “insane” after watching EliseAI grow from a narrow 2021 leasing product into broader residential operations by 2024. The founders found the administrative structures surprisingly similar: staffing shortages, regulation, repetitive calls, structured intake, scheduling, and bloated costs passed to consumers.
Song accepts both sides of the healthcare-cost puzzle: cheaper and better care may lead people to consume more, which can improve longevity and happiness, but “I don’t think we’re getting a better admin experience.” Administrative costs have risen faster than clinical costs because phone-based, unstructured interactions defeated earlier software. She does not think prior healthcare technology adoption was a large enough boom to have reached consumers yet, but is hopeful AI will lower costs, improve outcomes, or both.
The platform could extend from first contact through billing and post-appointment communication. Today a patient spends ten minutes with a doctor, receives instructions, and it is “good luck from there onwards”; AI reminders could improve adherence, provide more time for questions, bridge language barriers, and include family members who missed the visit.
Song’s retrospective is that she would have started with affordable housing: it combines every industry problem with maximal compliance, paperwork, administrative drag, and slow adoption. That now informs the healthcare strategy—start with the most underserved and complex—while pursuing the ultimate outcome of making housing and healthcare cease being cost concerns for the average person.