Pioneering the AI-rollup model in the property management space with Dan Lifshits, Co-Founder @ Dwelly
Dan Lifshits reveals how Dwelly is pioneering the AI-rollup model by acquiring traditional property management companies and forcing operational standardization—a critical difference from failed buy-and-build approaches. He demonstrates how targeting hard-to-reach landlords, initially seen as a weakness, became their moat, and explains why most AI researchers fail at building practical applications without deep operational understanding.
Key takeaways
- •AI-rollups succeed by forcing acquired companies to standardize processes rather than adapting technology to existing workflows—the opposite of traditional buy-and-build models.
- •Customer acquisition challenges can become competitive advantages when they create barriers that prevent competitors from easily replicating your approach.
- •Data-driven process optimization often means abandoning industry best practices entirely rather than simply selecting the best existing option.
- •Technical AI talent without operational understanding consistently fails to build practical applications that work in real business environments.
- •Standardized processes must be established before AI implementation—AI should facilitate state transitions, not create the operational framework.
The essay
Property management companies have been trying to crack the landlord market for decades. Most fail because landlords are notoriously hard to reach , they hide behind shell companies, avoid marketing channels, and resist new technology. Dan Lifshits turned this weakness into Dwelly's core advantage by acquiring the letting agencies that already serve these elusive customers, then rebuilding them with AI from the ground up.
This "AI-rollup" model represents a fundamental shift in how technology companies should think about market entry in traditional industries. Instead of fighting for direct customer acquisition, Dwelly buys established letting agencies and forces them to adopt standardized, AI-driven processes. The approach sidesteps the customer acquisition problem while creating something much more valuable: a data-driven view of what actually works in property management.
Why Traditional Rollups Fail Where AI-Rollups Succeed
The difference between Dwelly's approach and typical buy-and-build strategies comes down to standardization. Traditional rollups try to preserve existing operational practices to keep acquired teams happy. Lifshits argues this is exactly backwards. "You cannot do that while preserving the variety of different operational practices and having a very versatile product supporting these different ways of work," he explains. "You can achieve it only by" forcing standardization across all acquired agencies.
This sounds harsh, but it enables something impossible with traditional rollups: using data to identify optimal processes across the entire rental lifecycle. Dwelly doesn't just pick the best practices from acquired agencies. They analyze all existing approaches and often "select none of the existing and reviewing all of them completely and rewiring your new one instead to replace the old ones." The AI layer only works when processes are standardized enough to move systematically "from state a to state b whenever you predefine know which states you wanna process to follow through."
This creates a compounding advantage. Each acquisition provides more data about what works, which improves the standardized processes for all future acquisitions. Traditional rollups can't achieve this because they're stuck supporting multiple operational approaches instead of optimizing a single one.
The Implementation Challenge Most AI Companies Miss
The technical execution of AI-rollups requires a specific type of engineer that most AI companies don't have. Lifshits points out that "top AI researchers working on the fundamental models itself" often fail at building practical applications because "application layer and the fundamental layer are slightly two different things." Success requires someone who understands both the existing operational processes and how to "reengineer that with the most progressed existing technology."
This explains why so many well-funded AI companies struggle with real-world implementation. They hire brilliant researchers who can build impressive models but can't bridge the gap between AI capabilities and messy business processes. The AI-rollup model forces this bridge-building because you're not starting with a blank slate , you're taking existing workflows and systematically replacing human decision-making with algorithmic decision-making.
The Broader Lesson for Founders
The most important insight from Dwelly's approach applies far beyond property management. When investors or customers resist your model, the instinct is to pivot toward what they want. Lifshits suggests the opposite: double down on what makes your approach different, especially if it solves a fundamental problem in a new way.
"The only thing that keep you afloat whenever you're hearing 200 rejections of VCs" is "your personal conviction" that the difficult path is actually the right path. In Dwelly's case, targeting landlords through acquisition instead of direct sales seemed like a weakness until it became their core competitive advantage.
This framework applies to any founder building in traditional industries. Instead of fighting entrenched customer behaviors, acquire the intermediaries who already have those customer relationships. Instead of trying to work around existing processes, force standardization so you can optimize systematically. The path of most resistance often leads to the strongest moats.
The AI-rollup model will likely spread beyond property management as founders realize that acquisition can be faster than customer education, and standardization enables optimization in ways that flexibility never can. Watch for this pattern in other fragmented, relationship-driven industries where the customers are hard to reach but the service providers are hungry for better tools.
Two episodes. Free. Clips before your next meeting.
No card. No setup call. Paste your episode and see what Clypt surfaces.