AI will not remedy all our real estate woes

This winter, American realtors have a plethora of problems to ponder. Surging interest rates have created a frozen residential housing market and shifting working practices will push office vacancies to 55 per cent above their pre-pandemic peak by 2030, according to new research from Cushman & Wakefield. The prospect of stranded assets looms.

But if you listen to the chatter among realtors right now, there is another topic sparking even more debate: ChatGPT, the chatbot equipped with generative artificial intelligence launched last November by OpenAI, with backing from Microsoft.

The explosive spread of this tool has already sparked a noisy debate about the impact of AI on sectors such as education, science and the media. The real estate world, however, has attracted less investor focus.

That is a mistake. A wave of AI experimentation in this sector is already under way: brokers are using ChatGPT to write listings and social media posts, calculate mortgage payments, scour real estate databases and amass expertise on new fields, such as agriculture. As these experiments accelerate, they prompt a bigger question: will an AI-equipped realtor really need as many human realtors — and brokerage fees — in the future?

Or to put it another way, is rent-seeking in the property sphere about to come under attack from robots? Not in the literal sense of involvement in leases, but as described by economists: when powerful incumbents rig a market, by making it opaque or complex, to extract high fees.

Realtors themselves do not believe they will lose their jobs. “A common feeling [in the property sector] is that it will help rather than replace real estate professionals,” the Cushman & Wakefield report insists. “Some tasks will be automated [but] it’s more about helping to make people really good at what they do.”

Maybe so. Robots, after all, cannot sense the intangible qualities of a building, in a way that humans can. Nor can they read the body language of purchasers and sellers, and provide the necessary encouragement or reassurance. Given the extreme anxiety involved in many property trades, this is a serious shortcoming.

Moreover, the AI tools such as ChatGPT that are sparking such excitement do not always function well without human oversight. To understand the risks, consider an experiment recently carried out by Sarah Bell, an AI specialist focused on real estate. When she asked ChatGPT to evaluate the desirability of different tenants based on nationality, she discovered that the python computing code inside the bot was profoundly prejudiced against Australians.

More alarming still, it was impossible to determine exactly why it disliked Australians, because the system was a black box. A real estate principal who unknowingly deployed the anti-Australian python code would be “responsible” for violations of anti-discrimination legislation, Bell notes.

What the industry really needs is not so much AI in the sense of artificial intelligence, but another type of AI: augmented intelligence — ways to help humans to think smarter. The only way to achieve that is to add in a third AI, anthropology intelligence, which would imbue the process with vital cultural context.

Even if humans are needed to make sense of AI, it is not clear that the industry needs as many of them today. After all, in the past decade, the internet has already introduced new levels of transparency into the business: non-professionals can now use property websites to value properties, see listings, arrange mortgages and connect sellers and buyers.

Despite this, the number of residential real estate brokers and dealers in America (some 562,000 at last count) has not declined, but appears to be on the rise: the number is projected to expand by another 5 per cent over the next decade. Employment in the commercial real estate sector is almost 4mn — and is rising.

Neither has the new era of digital transparency dented sky-high brokerage fees. In 2020, average commission rates on residential property sales were 5.66 per cent. That was less than the 6.02 per cent level seen in 1992 — but higher than the levels recorded between 2011 and 2018. This is utterly perverse. Or, as an economist might say, a strong sign of rent-seeking.

Might this now change? Realtors obviously hope not. But entrepreneurs are sensing an opportunity for disruption, particularly given the other macroeconomic pressures now bearing down on the market. “Technology entrepreneurs have a unique advantage to start a real estate tech company in AI today [since] most incumbents have had a challenging last two years,” argues Kunal Lunawat, one such wannabe challenger.

Investors — and economists — would do well to watch what happens next, not least as a test case for whether tech innovation can challenge rent-seeking. After all, if robots can make the industry more efficient and bring down commission rates, many property owners will be thrilled. Rightly so: with or without AI, the sector is overdue for reform.

gillian.tett@ft.com