The AI Shift in Proptech: Why Horizontal Models Are Eating Vertical Software

We track over 2,500 real-world problems in construction, property management, and real estate. When Brendan Wallace says the definition of proptech is expanding, he's not just talking about software categories—he's describing a funding shift that our data confirms in unexpected ways.

Wallace, co-founder of Fifth Wall, recently told Crunchbase News that horizontal AI solutions are starting to dominate vertical proptech tools. He points to Higgsfield, a visualization platform originally aimed at film, which Fifth Wall backed for real estate applications. It's a compelling bet: why build a custom rendering tool for unbuilt homes when a general-purpose AI can do it better?

But our data tells a more nuanced story. While horizontal AI is undeniably powerful, the most severe problems in the built environment are often the least glamorous—and the most resistant to generic solutions.

Where the real pain lives

Take contractor payment disputes. It's a problem we rate 5/5 in severity. A contractor finishes a job, but the property owner withholds payment, claiming the work was subpar or incomplete. Without video evidence or a clear paper trail, the contractor is out thousands of dollars. It's a legal and compliance nightmare that a foundation model can't simply "understand away." The solution requires industry-specific knowledge of lien laws, local regulations, and dispute resolution workflows.

Property management has similar high-severity issues. We've logged 864 problems in that sector alone, and the top opportunities often involve distressed asset management—where owners are underwater, tenants are at risk, and regulatory compliance is non-negotiable. A general AI can draft an email or summarize a lease, but it can't navigate the eviction process in Cook County while respecting fair housing laws and communicating empathetically with a struggling tenant. That takes a hybrid approach: horizontal AI muscle paired with deep vertical expertise.

Wallace himself acknowledges that construction tech is "probably the biggest graveyard in all of venture." Our data backs that up. We track 753 problems and 455 app ideas in construction, but the landscape is littered with point solutions that didn't gain traction. Why? Because they often addressed surface-level inefficiencies rather than the underlying high-stakes failures. Recurring leaks, accounting overwhelm, and payment disputes are systemic, not cosmetic. They require domain knowledge and trust—two things a startup can't fake with a wrapper around GPT-4.

The horizontal wave is real, but it's not the whole story

There's no denying that foundation models are getting better at adapting to specific use cases. Wallace's point about ChatGPT 2023 vs. Claude today is well taken. The capacity for memory, personalization, and natural language instruction is improving at a pace that few predicted. For simple applications—tenant engagement apps, basic lease summaries, even some visualization tasks—horizontal AI will likely commoditize what were once vertical niches.

However, our data suggests that the most valuable opportunities are where horizontal AI alone is insufficient. These are problems with severity scores of 5/5 and rising trends. Distressed property management, for example, requires not just AI but also financial restructuring expertise, local market knowledge, and often a physical presence. That's why Wallace's push toward "full-stack real estate businesses" makes sense: the venture-scale opportunities are increasingly in owning and operating assets with technology as a differentiator, not just selling software to asset owners.

What this means for indie hackers and small teams

For indie hackers, the takeaway is clear: don't try to compete with OpenAI or Anthropic on general capabilities. Instead, find the narrow, high-pain problems that horizontal AI can't solve well. Compliance tracking for small landlords. AI-driven mediation for contractor disputes. Tools that help property managers navigate local eviction laws while reducing legal risk. These aren't glamorous, but they're sticky, and they solve problems that people will pay to make go away.

One insight from our data: a real estate agent lost a high-value listing because they couldn't recall specific CMA data under pressure. The proposed solution—an AI coach that helps agents practice and recall key points—is a niche application that horizontal AI alone wouldn't address. It's about human performance and confidence in high-stakes moments, not just information retrieval.

Watch the compliance angle

Another area Wallace didn't touch on is legal-tech for real estate. Contractors losing payments and property investors suffering from negligent managers point to a need for tools that protect parties through documentation, evidence capture, and automated compliance. These problems are severe and frequent enough to rank high in our opportunity scores. A startup that can automate lien filings or provide AI-powered audit trails for property maintenance could carve out a defensible niche.

The bottom line

The proptech landscape is shifting, and Wallace's observations are directionally correct. Horizontal AI will absorb a lot of the simple, replicable use cases. But the built environment is messy, regulated, and emotionally charged. The founders who win will combine the power of foundation models with the kind of domain depth that can't be replicated by a general-purpose tool. That's where the real opportunity lies—for builders and investors alike.