This is a public post, because it is me simply musing on some developments and ideas around AI. The hype is real, after all.
A post on Inman promoting yet another “artificial intelligence-powered real estate search platform” is what sent me down this path. I’m certain that HouseMe.ai is very cool, and the fact that the co-founders are three luxury agents instead of Silicon Valley tech bros is a good sign. But reading through that article made me wonder if real estate is not pursuing the wrong goals with AI and not yet quite understanding what AI is good at doing and what it’s less-good at doing.
So while I will reference the write-up in Inman, this is by no means a review of HouseMe.ai or a criticism of it at all. HouseMe.ai is merely the jumping off point for considering ideas I have knocking around my head, because what it claims to do is something I have noticed a lot of “AI tools” in the industry also attempt to do.
Lifestyle Search
I clicked into the article because it mentioned “lifestyle search” right in the title. And that’s a topic I’ve engaged with in the past, long before AI was a thing. From the article:
“No one has ever been able to search for real estate this way,” Coombe said in a statement. “This is a fundamentally new category, using AI and real data to let buyers, sellers and agents customize their search around exactly what they need.”
Coombe added that “buyers don’t just want a listing anymore, they want to search for the lifestyle they seek.”
This has been a recurring theme in real estate for ages now. In fact, I was personally involved with something called “lifestyle listings engine” back in 2009. The idea then was the same: buyers don’t want houses, but a lifestyle. Then over the years, there has been numerous efforts to offer lifestyle search by a lot of companies like planetRE, Local Logic, URBAN4M, and others.
I don’t know that “lifestyle search” ever really worked out like people thought it would. Turns out, consumers do kinda know how to search for the lifestyle they seek and are now just looking for properties that fit it. Remember that most people move within 15 miles, so they already know the area; they’re just looking to upgrade the home. So very few people are going to a computer and asking, “Where should I live?” They already know; they just want to know what’s for sale.
Of course there are gaps, especially if you’re moving into a new town or state; ask me how I know. But… if there are those gaps, I don’t know that people want to go ask a computer.
In my case, when I first moved to Las Vegas, we rented for a year specifically because we didn’t know what it was like to live here, and didn’t know what areas and neighborhoods would fit our lifestyle. We asked agents for ideas, but even then, we wanted to experience it firsthand before buying a house.
In other cases, maybe people can’t rent while living/looking so they really need guidance on where to live and what house to buy based on their lifestyle. In such a case, I submit to you that AI is the wrong thing to ask. A human being is the right fit for purpose.
Limitations of AI
I am no AI expert. I just vibe code a lot. And use AI every single day.
Over time, my experience with AI is that it is wonderful at some things, but really not that good at other things. For example, AI is really not good at pushing back, at disagreeing, and at correcting mistakes. I think all of them are programmed to be Yes Men. I literally have to tell my AI to stop kissing my ass.
Frontier AI like Claude Fable 5 or GPT 5.6 also tend to overcomplicate shit and overthink everything. That’s a great trait when you’re looking to code up a web app, but it’s not that great for something like, “I enjoy biking; where should I live?”
I literally just tried this experiment. I asked Kimi K3 (near-peer to Fable 5) where I should live in the Las Vegas Metro area if I enjoy competitive shooting. Which I do.
It did a very very respectable and impressive job. It identified three main shooting ranges that host competitions, mentioned other quality-of-life factors like state gun laws, no state income tax, even the availability of BLM land. Very respectable. Then it asked me a bunch of questions to “clarify” things so it can help me more.
However, the AI did not know some important details that affect competitive shooters like me. For example, it left out Pro Gun which hosts a match every weekend, or ease of getting to highways for travel to Arizona or Utah (St. George has a very active scene). It ignored how difficult it is to get memberships at some of the private ranges. It did not know anything (how could it?) about the quality of competition across the Las Vegas valley.
You know who would have known all of those things and could have advised me far better for my “lifestyle search”? A real estate agent I know who happens to be a competitive shooter.
This is just a limitation of AI. It has been trained on a ton of data, and it knows a lot about a lot of topics. And maybe in the future, with some super AI that tech bros are dreaming up, we’ll get the real-deal Wintermute. But today, AI isn’t great at “lifestyle search” because it is not a human and does not have a lifestyle. It can offer fact-based “advice” but miss out on so much that is important to a person’s actual lifestyle. Because it isn’t a person.
I genuinely don’t think this is because my hobby is such a unique and weird niche. If I asked Kimi or ChatGPT or Claude about fine dining, it can look up all the reviews, look at stars on Yelp or Google, search through forums and such… but I do not believe AI can come close to a human who loves fine dining. Because Claude has never enjoyed xiaolongbao or dined at Sparrow + Wolf.
As I said, I work with AI every single day. It is great at some things. It’s not great at others, and the thing it is most not good at doing is being human. Lifestyle is inherently and essentially human.
The “Intelligence Report”
Now… one thing AI is really good at doing is compiling a huge amount of information and data and generating reports. So I found this passage very interesting:
HouseMe.ai searches also generate an “Intelligence Report.” At its foundation is the True Cost Calculator, which breaks down the full cost to close. The AI Investment Thesis builds on that with broker-level analysis of each property, including financial summaries, key highlights and a plain-language read on potential risks.
This, AI can do and do very well.
The report also includes the AI Valuation Score, a public 0-10 fair-value rating assigned to every active listing. Paired with a Negotiation Strategy tool that generates data-backed offer recommendations based on days on market, comparable sales and local conditions, HouseMe.ai says it gives buyers an edge in how they evaluate and approach each opportunity.
Valuation Score? Negotiation Strategy tool? Data-backed offer recommendations eh?
Rounding out the report is the Area Market Pulse, with neighborhood data on pricing trends, inventory and market conditions, plus a conversational AI interface that answers questions about any listing in 97 languages.
Neighborhood data on pricing trends, inventory and market conditions that can be delivered over “conversational AI interface” that answers questions, eh?
So if HouseMe.ai can do all of that… I have a question for the buyer agent that HouseMe.ai will eventually connect me to: What would you say you do around here?
The AI can give me a True Cost Calculator, and give me a deep analysis of each property, including financials, and potential risks. It can help me negotiate with Negotiation Strategy that generates data-based offers. And AI can give me market conditions, inventory, and answer all my questions.
What do you do? Open doors? What if the home has a smart-lock that the AI can work with the homeowner or listing agent to let me tour it? When we are touring, what can you tell me that the AI hasn’t already told me? Once we are done, and I want to make an offer, what is that you do when the AI is giving me the full negotiation strategy and data-backed offers?
I know my friend Greg Robertson would say, “AI can’t smell the cat.” But I can. I don’t need a human buyer agent to smell the cat for me.
Oh, that’s right… real estate is an intensely stressful and emotional transaction and you will be there to console and soothe me like a psychologist, although you are not licensed to be one and certainly not qualified to be a psychologist. Why don’t I just pay my actual licensed therapist her $150/hr rate to console me and hold my hand instead of paying you 3% of the purchase price?
For that matter, did you know that 1 in 4 adults use AI for mental health? From the Stanford study:
A striking 24% of surveyed participants use LLMs for mental health. These users are more likely to be young, male, and Black and have poorer mental health and quality of life. They report difficulty accessing traditional mental health treatment – particularly due to cost and insurance coverage – and use LLMs because they are free, convenient, and available when needed. Users rely on LLMs for social and emotional support, to learn therapy skills and tools, and to supplement existing therapy. Non-users of LLMs for mental health expressed doubts about LLMs’ empathy and trustworthiness. Our sample likely overrepresents technology adopters, conservatively adjusting for this based on address-based estimates of population LLM use suggests that 13-17 million US adults may use general-purpose LLMs for mental health.
If people are willing to treat Claude as a therapist for mental health, you don’t think they’ll talk to Claude about the anxieties from trying to buy or sell a house?
So… what would you say you do around here, Mr. or Ms. Real Estate Agent?
Combining the Above
I am not an AI doomer. I don’t actually believe that AI will replace all human workers. And combining the two above suggests why.
AI is not good at doing “lifestyle search” because it fundamentally does not understand what “lifestyle” is. It doesn’t have a lifestyle and has no experience with it. (Again, maybe future AGI type stuff may make this statement outdated, but then I just hope to merge with Wintermute and not care about housing or food or anything.)
AI is really good at generating reports and recommending data-backed offers and such.
The solution is obvious: let AI do what it’s good at doing, then hand that off to the human who is good at having a lifestyle, at dining out, at having hobbies and families and pets, at you know, being human.
The effort to “be more efficient” to the point of having consumers chat with AI in 97 languages is bad for two reasons. One, it truly devalues the human real estate agent, and two, it delivers a sub-par experience for the consumer. As HouseMe.ai said in Inman:
HouseMe.ai emphasizes that it was founded by working luxury brokers, not outside technologists. The company says that grounding in the day-to-day realities of high-stakes transactions shaped a platform built to answer the questions buyers and sellers most frequently ask.
I end up wondering why you would want the platform to answer questions buyers and sellers ask. Wouldn’t it make more sense instead to have a human being informed by AI answer those questions? But if a human did answer those questions, what value is the person adding there? If she is simply repeating what the AI did, there is no value. So what is the human agent adding?
What is it that you do around here?
What we have to figure out, along with every industry in America today, is to specialize better and let AI do what it’s good at while humans do what they’re good at, and then train the humans to be ever better at those things. Because the AI is nonstop training to be better at what it’s good at, and trying to get better at things it’s not good at today.
Seems to me that it is simply a matter of time before somebody somewhere tries to operate a brokerage that has no human agents and just uses AI to help a buyer or seller do a transaction. That’s coming, and I am confident of that.
I am equally confident that that brokerage will carve out a niche and no more. Because AI isn’t good at some things. What I don’t know today is what value the human real estate agent will add. We had best focus on figuring that out as quickly as possible.
-rsh