Ben Miller, CEO of RealAI and Fundrise co-founder, argues that real estate's AI problem isn't about better language models. The industry sits on mountains of untapped data, from maintenance issues like leaking toilet flappers to tenant demographics, yet still relies on manual spreadsheet analysis.
The missing ingredient is a proprietary data layer. General-purpose AI tools cannot access the specific, structured information that real estate operators generate daily. Buildings produce continuous streams of operational data, but this information remains siloed within individual property management systems, accounting software, and tenant records. Without consolidating and standardizing this data, even the most advanced AI models struggle to deliver real value.
Miller's insight cuts through the current AI hype cycle. The real estate industry doesn't need ChatGPT or another large language model. It needs tools designed around actual workflow problems. A property manager analyzing maintenance costs across a portfolio, a developer evaluating neighborhood trends, or an investor assessing portfolio risk all need domain-specific intelligence built on reliable data, not generic language capabilities.
RealAI appears to be addressing this gap by building infrastructure that connects disparate real estate data sources. The company recognizes that competitive advantage in AI comes from owning proprietary datasets and understanding how to process them for specific use cases. This aligns with how successful AI companies operate across other industries. Stripe's fraud detection doesn't win because of better base models but because Stripe controls payment transaction data at scale.
For the broader real estate sector, the implication is straightforward. AI adoption depends less on adopting new software and more on data standardization and integration. Real estate firms that can organize their operational data will extract value from AI tools. Those stuck in spreadsheets will fall behind not because AI isn't advanced enough, but because they lack the foundation to use it.
Miller's framing also suggests that venture opportunity in enterprise AI isn't necessarily