Your brokerage needs an AI policy, and it should fit on one page

Ask a broker-owner what their AI policy is and you get one of two answers. Either they are waiting on counsel to finish something, or they say there isn't one yet, usually followed by a nervous laugh.

Meanwhile a hundred agents in that brokerage are pasting client financials into free chatbots, generating listing copy nobody proofreads, and running comparative market analyses through tools that make up numbers. The absence of a policy is not neutrality. It is a policy, and it is the worst one available.

Why the long document fails

The twenty-page version fails for a reason that has nothing to do with its content. Agents are independent contractors with full days who will read a document exactly once, if that. Anything longer than a page becomes a thing that was distributed rather than a thing that changed behavior.

The other failure is that long policies try to enumerate approved tools. That list is wrong within a quarter. Write principles about data and accountability, which stay true, not a catalog of software, which does not.

A policy nobody can recall under pressure is decoration. Write the one your agents can remember in the car.

The one page

Six sections. Plain sentences. This is the skeleton I give brokerages and it adapts to nearly any size.

1. 1What never goes into a tool. Client names, addresses tied to names, financial documents, Social Security numbers, contracts, anything from a transaction file. State it as a bright line, because bright lines are the only kind people remember.
2. 2What is fine. Drafting, summarizing your own notes, rewriting your own copy, market research, learning. Say this explicitly, or cautious agents will assume everything is forbidden and you lose the upside.
3. 3The accountability sentence. Anything you send under your name is yours. AI is not a defense for an error, a misrepresentation or a fair housing violation. Read it before it goes out.
4. 4The verification rule. Any number, date, legal statement or property fact gets checked against the source before it leaves. Full stop.
5. 5Disclosure. Where you tell clients AI was involved and where you don't. Being clear here prevents the awkward improvisation that damages trust.
6. 6Who to ask. One named person, and an explicit statement that asking is encouraged and never held against you.

Section three is the one that matters

If you strip everything else, keep accountability. The entire category of AI risk in real estate collapses into one behavior: someone sent something they had not read. The fabricated statistic in a market update. The generated description claiming a permit. The chatbot answering a fair housing question it had no business answering.

None of those are failures of technology. They are failures of review, and a one-line rule addresses them better than fifteen pages of definitions.

Roll it out in a meeting, not an email

Emailed policies are archived unread. Take fifteen minutes at a sales meeting, read the page out loud, and then do the part that makes it stick: ask what they are already using. You will hear about six tools you did not know were in the building. That is the real inventory, and it is worth more than any audit.

Then revisit it once a quarter for ten minutes. Not a rewrite. A check on whether anything on the page has become wrong. A living one-pager beats a definitive document that ages in a drawer.

The Takeaway

One page, six sections, principles rather than tool lists, and one unbreakable rule: whatever goes out under your name is yours, read it first.