Few things in the tech world get people more excited than artificial intelligence. AI research has been around for decades, but we’ve seen endless hype about it in the past few years. Elon Musk made headlines (as he always does) in 2018 when he said super-intelligent AI poses the greatest existential threat to humanity, more so than climate change and nuclear weapons. Various studies and reports lately claim AI will massively boost the economy, followed by the predictable headlines that robots are coming for our jobs.
But how much of this talk is hype, and how much is based in reality? For every AI enthusiast, there’s a scientist and researcher out there who will be quick to tell you that we’re still far away from the holy grail of artificial general intelligence. As far as AI replacing human intelligence, some experts like Erik J. Larson, a writer, tech entrepreneur, and computer scientist, say it’s essentially a myth perpetuated by sci-fi films. The complexity of the human mind is much harder to replicate than many anticipated.
There has also been a fair share of headlines about a coming AI revolution that’ll disrupt commercial real estate. AI is already being used in several ways in the real estate world, like lease reading and data entry. The tech has great potential to do more, but a ‘revolution’ may not be coming. And if it does happen, it’s probably further down the road than we’ve been led to believe.
Bar Mor, CEO and Co-Founder of Agora, knows a thing or two about this. Mor grew up in the real estate world with his father being a real estate entrepreneur. He always wanted to invent something that would change the way he saw his father do business. For a long time, Mor assumed the tool would be artificial intelligence. But after much observation of the industry, he changed his mind, determining the future of real estate depended more on automation.
Commercial real estate, in his mind, isn’t yet ready for an AI takeover. Instead, Mor created Agora, an investment management software that leverages automation and data analytics. “The main reason AI hasn’t been more widely used is the data,” Mor said. “A lot of things in commercial real estate are still done manually, information is put in Excel and in many fragmented databases. So, first, the information needs to be digitized. Only afterward, when the AI can be fed this data, will artificial intelligence have better applications in real estate.”
Learning machines
A simple definition of AI is technology programmed by humans that uses algorithms, logic, and data to make better-informed decisions. Machine Learning is a branch of AI where a machine is given access to data to learn based on past experience and historical information. In machine learning, algorithms build a model based on data to make predictions or decisions without being explicitly programmed by humans to do so. Then there’s Deep Learning, a type of machine learning based on artificial neural networks that support unsupervised learning from data sets.
AI is being used in some narrow ways in commercial real estate already, mostly in gathering, interpreting, and analyzing large sets of data. For example, automated property valuation models (AVMs) gather data about site locations, transportation access, and other statistics like demographic trends to estimate property values. AI is also helpful in doing deep market analysis to identify rental trends, occupancy rates, and usage statistics within particular geographic areas. Perhaps the best use of AI in commercial real estate today is collecting data through building automation and sensors.
All these advancements are great, but people like Mor think the possibility of AI replacing real estate agents and brokers one day is far-fetched. When I spoke to Mor, he told me the most optimal use of AI will be helping real estate professionals in more of an assistant role, streamlining their day-to-day operations, and letting them focus on the parts of the job that require human ingenuity and interpersonal skills. “I think commercial real estate will always stay based on relationships between people, but AI will be a tool used to leverage results,” Mor said. “AI tools will empower companies and individuals to do their jobs better.”
In this way, Mor is like many others who see AI complimenting people in the real estate world rather than replacing them. A good example of this is the use of chatbots by real estate agents today. Twenty-eight percent of realtors use chatbots now, and the real estate businesses profit more from the tech than any other industry, according to Brillio, an IT Services provider. Chatbots enable agents to bring in leads sometimes more effectively than humans, and consumer attitudes toward them have become more positive over the years. Chatbots provide quicker responses to simple questions, which can free up time for more important work. But would some version of a chatbot replace an agent? Probably not. This tech streamlines operations, but buying and selling a house is a very personal experience most wouldn’t want to do with a robot.
Too much zest
A recent example of the limitations of AI in real estate is Zillow’s iBuyer model, which the company discontinued in late 2021 after sustaining substantial financial losses because of it. The iBuyer model was based on buying homes directly from sellers and then reselling after minor repair work. Much of the model depended on the ‘Zestimate,’ a machine-learning-assisted estimate of a home’s value based on reams of data like property and tax records, pictures of homes, and homeowner-submitted data. After determining the Zestimate, the company would do an in-person evaluation, figure out the number of repairs needed, and then make a final offer.
See also
Zillow bought tens of thousands of homes through the Zillow Offers idea, but they sold far fewer than they purchased. The biggest challenge Zillow faced with the venture was accurately forecasting the future price of its homes three to six months out, something their machine learning system wasn’t quite good enough to do. The company’s AI could process vast amounts of data, but what if a real estate agent picked out a critical valuation factor that didn’t appear in the database? Zillow has spent years improving their valuation model, what they call a ‘Zestimate,’ a vital part of their brand. But the Zestimate has a median error rate of 1.9 percent for homes on the market, according to Zillow spokesperson Viet Shelton. Being off by 1.9 percent on a home worth $500,000 comes out to nearly $10,000.
Using AI for something like valuation modeling can produce some reasonably accurate calculations. But using those estimates to make real-world decisions, especially at a large scale, proved disastrous. This is because not everything that goes into a property’s valuation gets captured in the data. For example, if the Zestimate misses a hidden problem like a crack in the foundation, the value drops significantly without the algorithm taking it into account. Many intangibles in home buying also can’t be captured in data, such as sentimental value or if a buyer’s relatives live in the neighborhood. Mike DelPrete, a real estate technology strategist and scholar-in-residence at the University of Colorado Boulder, told CNN the Zestimate is more so a ‘toy’ to pique your interest when looking up home values. He emphasized it shouldn’t be thought of as a way to accurately predict home prices now or in the future. Mor, the CEO of Agora, added there’s a lot of data Zillow’s AI was probably missing, such as not capturing off-market deals.
All this goes to show that, while AI can be a great tool in the real estate world, it’s still not at the level many would like. If AI does somehow replace humans in real estate, it’ll likely be in mundane jobs that involve data entry and tedious manual processes. Otherwise, real estate is a business centered on human relationships and intuition, which will never change. AI and machine learning can help us sort through and organize data, giving us insights into market trends and helping us make better-informed decisions. But as Mor said, before AI can really be launched at a larger scale, much of real estate’s fragmented data still sitting in Excel spreadsheets need to be digitized and collected in central databases.
The failure of Zillow’s iBuyer program shows that AI can provide reasonable estimates for property values, but there’s still an element of human touch that makes the difference. Algorithms and machine learning can point us in the right direction, but relying on them too heavily could be a mistake. Despite all the hype about artificial intelligence, many researchers know the complexity of the human mind is tough to replicate. The same goes for real estate, where the complexity of assessing value can’t easily be taken over by a machine.