Cautionary Tales: AI Only Works With HI (Human Intelligence)

 Artificial intelligence is making an impact – both good and bad – in industries all over the country.  

Here in Pennsylvania, the State Board of Medicine has filed an action against a company called Character.AI, claiming that user-generated AI “characters” carried on conversations with users in which the characters held themselves out as licensed medical professionals. As described in the filing, a character named “Emilie” not only suggested that the user (a state investigator) might have been suffering from depression, but also provided a (made-up) educational and license history, including a (made-up) Pennsylvania license number. The company is being charged with practicing medicine without a license.   

What does that have to do with the real estate industry? 

As a real estate licensee and a Realtor®, you are responsible for all of your work product, regardless of where it comes from and how it’s developed.   

If your web developer makes your brokerage name too small on your website, you are the one who gets fined by the State Real Estate Commission. If an unlicensed assistant starts giving real estate advice to consumers calling your office, you are the one who will be held responsible for that unlicensed activity. And if AI goes off the rails by improperly editing photos, giving real estate advice or telling consumers that it’s you, you are the one who will be sitting at the hearings.   

At every turn, real estate brokers and agents are being bombarded by AI-everything, promising to revolutionize your workflow, polish all your marketing, step up your customer service and just generally make things better in every way. Used judiciously, in conjunction with HI (human intelligence), providing human review and oversight, these goals might be achievable. But if you simply use and rely on AI output without adding in the HI, your professional reputation and even your license could be at risk.  

Consider these (based on real life) examples:  

• Agent feeds photos and property data into an AI model and asks it to write a property description that will maximize consumer interaction. The AI searches available data and finds that ads for four-bedroom homes with renovated kitchens get the most interaction, so it writes a description for a four-bedroom home with a renovated kitchen. Except that the property is actually a three-bedroom Cape Cod with a kitchen that hasn’t been touched since 1987.  
• Agent puts some dreary mid-winter photos into an AI editor and asks it to put in bright, appealing backgrounds. The resulting photos erase the power lines running across the property and insert a beautiful sunset behind the North side of the house (hint: the sun does not set to the North).   
• AI chatbot deployed on a broker’s website responds to a consumer’s questions about the real estate market by introducing itself as the broker, asking about their property wish list and budget, then recommends certain listings — all without any licensee review … or a Consumer Notice.  

A licensee who just blindly relies on these AI resources without a thorough HI review could easily find themselves on the wrong end of both ethics and licensure complaints. And your brokers may be there with you, after collecting the dreaded “failure to supervise” charge from the State Real Estate Commission. “I just assumed it was right because it’s AI” will not be a defense, so … do better.   

In all of those scenarios, the most obvious common-sense protection for licensees is to review the work product that’s being provided to ensure compliance. You can start by asking a few easy, common-sense questions that will give you a good start to protecting yourself.  

“Would I let an unlicensed assistant do/say that?”   

If work is appropriate for an unlicensed assistant, then it is probably suitable for a higher degree of automation (though still with a licensee review). But work that requires a license is either unsuitable for AI, or it should only be used with a deep and thorough review by the licensee.   

For example, an unlicensed assistant or AI could certainly write marketing copy, with the licensee reviewing and approving it before publication. But an unlicensed assistant could not tell a buyer what terms to include in their offer, so it’s probably not wise to use AI to try figuring out how to draft contract terms. Or even if there’s a place for it, the agent should still have their own conversation with the buyer to affirm those choices and explain what they all mean.  

“Can I effectively supervise AI?”  

Brokers (hopefully) have various methods to check up on their agents to be sure they’re following office policies and the law, and they could terminate those who do not. But how do you supervise AI?   

 If it’s writing copy, review the copy to make sure it’s accurate. If it’s editing photos, compare them to the originals and make sure they still offer a “true picture” of the property. If it’s an AI bot on your website … well, how do you know what questions it’s asking, what answers it’s giving and what information it’s collecting? If the answer is “I just have to trust that it’s okay,” then you’ll be quite exposed if some sort of complaint gets filed. Just like any other tool or product you use in your business, work with vendors to ensure you’re able to appropriately limit the AI capabilities to stay within the law and ensure that you have some way to evaluate its work to be sure that it’s not going outside of legal and ethical lines.