Hot take: You might need machine learning, not AI

“Machine learning is where I would start, specifically around questions around underwriting and valuations.”

That’s Garret Van Parys on why real estate companies may be overlooking one of the most practical applications of data science while rushing toward AI.

Garret Van Parys is the Chief Strategy and Analytics Officer of Esusu, a billion-dollar fintech platform turning rental payments into credit-building outcomes. He previously spent more than thirteen years at Invitation Homes, most recently as Senior Vice President, Enterprise Analytics & Revenue Management, overseeing $3 billion in annual revenue while helping change how residential real estate investment management uses data and analytics to drive performance, strategy, and operational efficiency.

Garret helped lead the two largest mergers and integrations in single-family rental history and was recognized as an IMN "Rising Star Under 35." He holds graduate degrees from Duke University's Fuqua School of Business in Quantitative Management and an MBA from the University of Arizona's Eller College of Management.

Jonas and Garret go back to the early days of single-family rental, when the industry was scaling rapidly but the systems and data infrastructure needed to run it simply did not exist. Their conversation moves from those early days of Excel and Access databases to machine learning, AI, revenue management, customer retention, and what real estate companies should actually be doing with these technologies today.

In this episode

Building the data infrastructure from scratch. Garret takes Jonas back to 2012, when he joined Colony American Homes as roughly employee number 30. The company had an Access database and some Excel files, but no ERP or established property management systems. With the business acquiring homes at an extraordinary pace, the team had to build the systems needed to track acquisitions, renovations, inspections, and readiness from the ground up.

The operational challenge of scaling single-family rental. Managing hundreds or thousands of individual homes is fundamentally different from managing a large multifamily property. Garret explains how the distributed nature of single-family rental created enormous operational complexity, with homes in different locations and different stages of rehab. Even knowing where every home was in the process could become a major challenge.

How data helped make massive mergers possible. Garret was involved in both the Colony and Starwood merger and the later Invitation Homes merger. He explains how data and analytics became essential to answering practical questions about how to run a combined portfolio, including staffing levels, market overlap, operational efficiency, and organizational structure. The company went from zero homes to roughly 80,000 homes in about five years, making the ability to model and understand the business critical.

Why machine learning deserves more attention. Garret's central argument is that real estate companies may be moving too quickly toward AI without fully exploiting machine learning. He describes machine learning as more deterministic, less expensive, more straightforward, and generally more explainable. Once the underlying data is in place, he believes machine learning should often be the first step for solving fundamental problems in areas such as underwriting and valuation.

The technology is only as good as the people using it. Garret pushes back on the idea that an executive can simply open Claude or another AI tool and ask it to build a machine learning algorithm. The tools are increasingly accessible, but knowing whether an approach is scalable, durable, and actually solving the right problem still requires expertise. He compares these technologies to a paintbrush: having the tool does not automatically make someone a great artist.

Why business knowledge matters as much as technical expertise. Garret believes the strongest analytics teams combine technical capability with deep knowledge of the business. The people building models need to understand the real-world problem they are trying to solve, while also being curious enough to identify problems executives may not have even considered. He argues that companies should focus on finding and retaining people who can bring those capabilities together.

AI can uncover customer signals that humans could never monitor manually. Garret sees one of the most promising applications for AI in understanding customer interactions at scale. Recorded customer calls can be transcribed, analyzed for sentiment and topics, and monitored continuously. That creates a new ability to understand whether customers are satisfied, whether a representative handled an interaction well, and whether a customer may be at risk of leaving.

Retention may be the biggest ROI opportunity in rental real estate. Garret connects customer experience directly to economics. Every additional renewal can represent another year of rental income, and avoiding mistakes in critical customer interactions can have a meaningful impact on lifetime value. His argument is that keeping residents longer can be far more valuable than simply optimizing an individual transaction.

Stop treating 95% occupancy as a magic number. Garret challenges the idea that there is a universally correct occupancy target. The right answer depends on the relationship between rent, market conditions, turnover, and vacancy. Charging above-market rent may create more problems if it leads to longer vacancy and higher turnover. Garret explains why operators should think about the balance between rent and occupancy rather than blindly pursuing a particular percentage.

Vacancy is expensive, and the math matters. Garret recalls that vacancy at Invitation Homes cost roughly $80 per day. That changes the economics of chasing an additional $10 or $20 per month in rent if doing so means leaving a home vacant for several additional weeks. He argues that operators need to understand the tradeoff between price and occupancy rather than optimizing one metric in isolation.

The power of pictures in pricing. Garret discusses the potential to analyze everything from finishes and appliances to lighting, flooring, kitchens, and backsplashes. Different features may carry different value in different markets. As image analysis becomes more efficient, Garret sees an opportunity to incorporate this information into pricing models that traditionally relied on a much smaller set of variables.

Tread lightly before spending heavily on AI. Garret's advice to companies that are AI-curious is simple: you can always spend more money later, but you cannot take the money back once it is spent. He believes companies should think carefully about their infrastructure choices and avoid building expensive systems around assumptions about future AI costs and capabilities.

Start with efficiency and measurable ROI. Garret argues that companies should identify large problems that need to scale and focus first on efficiency. He does not recommend removing humans from a process unless a computer genuinely performs the task better and the people involved can then spend their time on more valuable, human-centered work.

Real estate needs a unified technology vision. Garret believes real estate companies risk falling behind more technology-oriented industries unless they develop a clearer view of how their platforms fit together. He points to Amazon as an example of a technology platform that can expand into new areas because its underlying infrastructure can support the change. For real estate, the question is whether technology platforms will eventually be flexible enough to support different asset classes without requiring entirely separate systems.

Use AI as a contrarian, not a cheerleader. In his personal life, Garret uses AI to challenge his own thinking. Rather than asking it simply to validate an idea, he asks where he is wrong, what could make the idea fail, and what opportunities he may have overlooked. He sees this as a way to strengthen an idea before taking it to a boss, partner, or anyone else.

Don't blindly trust AI. Garret also offers an important warning about using AI as a thinking partner. He describes AI as fundamentally a prediction model and points out that its tendency to be agreeable can turn the technology into a "sycophantic exercise." His advice is to use AI to challenge yourself, but never listen to it blindly.

Mentioned in this episode

Garret Van Parys · Esusu · Invitation Homes · Colony American Homes · Colony Capital · Starwood · Dwellsy · Yardi · Claude · ChatGPT · Amazon · UPS · FedEx · Zillow · AMH · Machine Learning · Artificial Intelligence · Revenue Management · Real Estate Analytics · Single-Family Rental · Multifamily · Valuation · Underwriting · Customer Lifetime Value · Occupancy · Vacancy · Pricing · AI Agents · Data Analytics

Net Effective is a conversation with the people running residential rentals about how they are actually using AI. New episodes weekly, about 30 minutes. Subscribe at neteffective.show.

Connect with Garret

Garret Van Parys
Chief Strategy and Analytics Officer, Esusu

Esusu