“[Artificial Intelligence] won’t replace real estate people. AI will replace real estate people who don’t have AI,” said Itay Ron, senior president of Northeast markets at Faropoint at the I.CON East conference this week in Jersey City, New Jersey. “Right now, we are reaching an inflection point where you can easily use data for commercial real estate,” Ron continued, discussing the state of artificial intelligence technology and how commercial real estate professionals can utilize AI in their investments.
AI in the Spotlight
Earlier this year, Microsoft-backed Open AI released ChatGPT to the public. This tool, which can create human-like text, showed the world the robust application of generative AI, but it’s not new. “AI has been used behind the scenes for many years. That release set off the arms race of AI between Microsoft and Google,” said Ron. And over the last year, you can see the impact of utilizing AI. News outlets and written content creators are being materially disrupted. For example, BuzzFeed, CNet and Insider are all news outlets decreasing their workforces by at least 10% and using AI to create news stories.
Insurance is another industry that is ripe for disruption, according to Ron. The industry is beginning to utilize active insurance; sensors on wearables will send your data to underwrite your health insurance policy. Sensors on your car will do the same for your auto. “By 2025, more than one million devices will be connected to sending data within insurance.”
AI in Commercial Real Estate Investing
Ron polled the audience about commercial real estate investment areas that could use the help of AI. Areas included underwriting, research, marketing, data and systems. He noted that firms are currently using AI in three ways:
- To streamline operations like drafting an offering memorandum.
- To analyze deals in aid of the underwriting process.
- To gain insights through data warehousing and business intelligence, where models mine your data to see if they can add insights or extract new value.
Ron said that Faropoint is doing all three. The firm focuses on urban infill investments and found that 76% of the urban infill sector is privately owned. This niche does not have data publicly available, and it’s a fragmented industry. While this presents a challenge, this is the very opportunity where AI can create value for those who can source the data behind these deals.
Implementing Big Data Solutions
According to Ron, the blueprint for implementing an AI model includes five steps.
- Digitization: Enter all desired data points into the computer.
- Collect and clean: Scrub the data, remove errors and create uniformity.
- Build models: Determine the problem you have, and try to solve it with integrity.
- Continuous testing/benchmarking/stability: Continually improve and test your models.
- Build the user interface: Make it easy for the business functions to use.
Ron said, “Eighty percent of AI startups fail to generate trust because they simply can’t explain the conclusions they generate. You need human oversight to look at the model and ask, ‘Does this make sense?’”
Real Estate Challenges and Opportunities
“We noticed that the challenges are a lack of data, which is quantity, and data integrity, which is quality,” said Ron. Deals are very complex, and you have to scrub the data to make it uniform, accounting for which leases are triple net versus gross, for example. Average rents were probably their hardest data point to tackle. And another aspect is that real estate comparisons age quickly.
Faropoint created a robust system for cleaning information before inputting it into the model. The company also made the largest real-time deal flow platform to capture new data. “You need a lot of data; this was a way to get it,” explained Ron. The firm then created a data lake consisting of historical property data from the past three decades, real-time data from their deal flow platform, and predictor data from demographics, labor, and other data points they layered into the model that affect the real estate industry. This process involved collaboration with 27 developers at the firm and required full support from all areas of the organization.
Further, in real estate, Ron explained, “the feedback loop is longer than for other industries like insurance. It can take several years to sell a property and realize the gain that an AI model would have predicted to be a good investment. For that reason, it can take a long time to know whether the model is correct.”
The result of the firm’s efforts is what Ron explained as the flywheel. Their model is deployed nationally at deal origination, helping the user underwrite while gaining user feedback to improve the model. It allows customization in its propriety investment platform, aggregates data to deepen its data lake, and creates new insights. User training and simulations further identify areas of improvement. The more the model is used, the more data it collects and the more accurate and precise it gets.
For a Company Just Getting Started
“AI is not a tactical move, and it’s not going away. It’s a tectonic shift with a net positive impact for our industry,” stated Ron. AI won’t replace your job but can make it easier and provide an excellent second opinion.
“If you are looking to improve your processes, there are already tools offered out there, and you will benefit more by looking into those instead of building your own from scratch. If you want to leverage your data to gain new insights, start capturing and cleaning your data yesterday. Make sure you collaborate because neither the AI developer nor the real estate guy can do it alone. And make sure you have total firm buy-in, even at the point of new hires.”
Ron said it is worth it for firms to create their models, and there’s no “winner-takes-all” setup. He did add, “The firm that creates the first automated valuation model will have the edge.”
This post is brought to you by JLL, the social media and conference blog sponsor of NAIOP’s I.CON East 2023. Learn more about JLL at www.us.jll.com or www.jll.ca.
