A fight over access

Last week, RealPage told its customers it was pausing all new site activations with EliseAI. Existing integrations stay live for now. New ones are suspended indefinitely.

It would be easy to read this as a vendor spat. I read it as a signal. EliseAI just raised $350 million at a $4 billion valuation in a round co-led by Andreessen Horowitz and Bessemer Venture Partners, and it says more than 6.5 million multifamily units use its leasing and resident communications tools. The flashpoint appears to be Apollo, an agent EliseAI launched last month that works across leasing, maintenance and operations rather than inside just one of them.

That’s the part investors should pay attention to. When AI only answered questions, incumbents could live with it. Once AI starts doing the work across systems, the system of record becomes contested ground. Fights like this happen when there’s a lot of value on the table, and real estate has more pressure to unlock that value right now than at any point I can remember.

The wall is a repricing, not a deadline

The maturity wall has been the industry’s favorite warning for three years. Lenders extended, borrowers waited for rates to fall, and the problem kept sliding forward. It isn’t sliding anymore. The Mortgage Bankers Association estimates $875 billion in commercial real estate loans mature in 2026, about 17% of the roughly $5 trillion debt market. S&P Global puts the figure closer to $936 billion and expects maturities to peak in 2027.

The bigger issue isn’t the due date. It’s the rate. A $10 million interest-only loan written at 4% costs $400,000 a year to carry. Refinance it at 6.5% and that becomes $650,000, before any change in valuation or leverage. Plenty of performing assets will need fresh equity just to stay put.

When owners can’t control the cost of capital, they look harder at what they can control. That means net operating income, the speed of underwriting and disposition work, and how much a team can handle without adding headcount. Operating leverage has moved from a nice-to-have to a requirement.

Consolidation now runs on data

The same pressure is reshaping who owns real estate. U.S. REITs announced four M&A deals worth $16.8 billion through mid-April, three of them take-privates, all targeting companies trading well below net asset value.

The deal that stood out to me was AvalonBay and Equity Residential. The combined company, Vivmark Residential, owns more than 184,000 apartments. When management explained the rationale, it didn’t stop at overlapping markets and lower overhead. It pointed to a proprietary dataset with more than 4 million lease transactions, 9 million service requests and 60 million customer insight data points, and to the technology and AI that dataset can power.

Vendors are making the same bet. RealPage bought Cherre this summer, and its CEO described the goal as a better data foundation for clients’ agentic solutions. Read that alongside this week’s news and the incumbent playbook gets clearer. Consolidate the data, then decide who gets to act on it.

Why the smart money is going vertical

Step back from this week’s headlines and the size of the opportunity comes into focus. McKinsey estimates generative AI could create $110 billion to $180 billion or more in value for the real estate industry. That figure isn’t a software budget. It’s labor, time and margin, which is exactly where Bessemer Venture Partners argues vertical AI competes. Vertical SaaS fought for a slice of IT spend, while vertical AI goes after the labor line of the P&L, and in an industry where so much work still runs through spreadsheets, inboxes and PDFs, that line is enormous.

Generalist investors have seen this movie before in the built world. Bessemer led an early $15 million round in Procore, the construction management platform, and owned 13% of the company when it went public in 2021 at a valuation above $8.5 billion. ICONIQ, its largest backer, held a stake worth roughly $3 billion at the IPO. Neither firm is a proptech specialist. They saw an industry with huge operational spend and outdated software, and they backed the company that went deepest on its workflows.

Now the same playbook is running with AI. Andreessen Horowitz led EliseAI’s Series E at a $2.2 billion valuation in August 2025. Thirteen months later, a16z and Bessemer co-led its new round at $4 billion, after EliseAI passed $200 million in annual recurring revenue and doubled revenue for five straight years. This week’s RealPage dispute is a real hurdle, and it may slow new activations. It doesn’t change why the valuation is that high. Investors are paying for a company that went deep in one industry’s operations and earned its place inside the workflow.

That’s the case for vertical over horizontal. Horizontal AI builds one capability for everyone, which puts it in a crowded fight with the model providers themselves. A vertical company that understands leases, fund structures and debt covenants competes on knowledge a general-purpose tool doesn’t have and can’t easily acquire. PitchBook reported that vertical applications overtook horizontal platforms in venture deal value and volume in the fourth quarter of 2025.

The capital is also disciplined. Proptech raised about $4.53 billion in the first half of 2026, roughly flat with a year earlier, according to the Center for Real Estate Technology and Innovation, and second-quarter funding fell 57%. But AI-focused proptech companies grew funding at a 42% annualized rate in 2025, nearly double their non-AI peers. Investors aren’t funding the category. They’re funding the companies whose software does the work.

The harder question is which of those bets will hold up.

Why most AI spend isn’t paying off yet

Adoption is no longer the hard part. Results are. Accenture found that only 23% of C-suite leaders report widespread, sustained value from AI across their organizations, even as AI spending heads past $800 billion this year. Real estate isn’t an exception, and in conversations with operators I hear the same three reasons.

The first is that most tools automate a task, not a workflow. AI can abstract a lease in minutes, but someone still has to check it, route it and update the system of record. The bottleneck moves without disappearing.

The second is the belief that data has to be cleaned before AI can be used. Real estate data is messy by nature. The same asset can show up three different ways in three systems, and every version is technically correct. Firms that wait for perfect data turn preparation into its own endless project.

The third is that generic AI doesn’t know the business. It can’t tell rent commencement from lease commencement or explain why a T12 matters. Without real estate logic and validation built in, small errors compound across every step, and in a business where one number can represent tens of millions of dollars, that’s disqualifying.

The deployments that work have one thing in common. They encode how the firm actually operates.

What smart investors are underwriting

If I were writing an investment memo on real estate AI today, I’d underwrite four things.

Execution over orchestration. A lot of companies now describe themselves as orchestration layers, and connecting systems matters. But routing information from one tool to another is table stakes, and this week showed how exposed a company is when its value depends on someone else keeping a door open. The durable value sits one level up, in an action layer that understands the data and takes the next step in the workflow instead of handing it back to a person.

Business logic and validation as the moat. Anyone can call a frontier model. Encoding a firm’s rules, checks and approvals so outputs are deterministic and auditable is much harder to copy, and it’s what makes AI safe to use where the numbers carry real consequences.

Software that compounds. When workflows carry a firm’s logic, they become the foundation for applications shaped around how that firm works. Those applications become proprietary assets that grow more valuable over time, instead of another subscription the team works around.

Operating leverage you can measure. With debt costs where they are, the right test for any AI investment is what it does to NOI and team capacity. Demos don’t refinance buildings.

This is the shift we’re building for at Outcome. Our AI action layer works from data as it already exists across a firm’s sources, from system exports and APIs to PDFs and spreadsheets, rather than depending on any single system of record. It automates workflows end to end, builds each firm’s business logic and validation into the workflows tailored to it, and uses that foundation to build the software a firm actually needs to unlock operating leverage, without the features nobody uses.

Who gets to act

For most of its history, real estate technology competed over who stored the data. The next round is about who gets to act on it. Capital pressure is forcing owners to find operating leverage now, consolidation is turning data into a strategic asset, and incumbents are already drawing lines around access.

For investors, that’s the opportunity. The companies worth backing won’t be the ones that move information faster. They’ll be the ones that turn a firm’s data and logic into work that gets done, reliably, every time.

Sources

• McKinsey $110 billion to $180 billion estimate: McKinsey
• Vertical AI and labor budgets: Bessemer Venture Partners, Building Vertical AI (January 2026)
• Procore early Bessemer round: Wikipedia (verify against a primary source); IPO stakes and valuation: The Real Deal, Pacific Coast Business Times
• Vertical overtook horizontal in Q4 2025 deal value and volume: York IE, citing PitchBook (verify against PitchBook)
• MBA and S&P maturity estimates: MaturingLoans.com (verify against MBA primary); S&P 2027 peak: RETS.ai
• REIT M&A, four deals totaling $16.77 billion: S&P Global Market Intelligence
• Vivmark Residential apartment count and dataset: Pulse 2.0
• H1 2026 proptech funding and PitchBook AI growth rate: AI for CRE Collective, citing CRETI and Bisnow; Commercial Observer
• Accenture 23% figure: Inman, Oct. 2, 2026