Real Estate Brokerage AI Policy: Why the PDF Is the Easy Part

Why the policy document is the easy part

A real estate brokerage AI policy is the easiest part of AI governance, because the hard part is that agents adopted AI before the brokerage chose anything. Most principal brokers already know they should have a policy. The harder problem is that the work the policy is supposed to govern is happening in tools the broker never picked, never saw, and can't open.

The scale of that gap is no longer a guess. NAR's 2026 REALTORS Technology Report, released September 22, 2026, found that 23% of agents use AI daily and 25% weekly, and that just 12% say they aren't using it and have no plans to. That is a statement about agents, not about brokerages. The survey doesn't say which tools those are or who chose them, and that is the point: the broker's view of them is thin.

Procurement used to be the supervision perimeter

For years, a broker's control over technology ran through the purchasing decision. The broker picked the CRM, the transaction system, the website vendor, and whatever sat in between, so the broker had at least read the contract. Victor Lund of WAV Group Consulting argues in a September 24, 2026 analysis that this perimeter has dissolved, and his plainest sentence is hard to improve on: "Agents are adopting general-purpose AI tools on their own."

Lund adds a second leak that gets less attention: vendors the brokerage already pays. In his words, "A feature that was not present when the contract was signed may appear in the next software update." So the exposure comes from two directions at once. Agents bring AI in through the front door of their own phones and browsers, and approved software gains AI features through the back.

Here's where we'd push past the usual advice. The common reaction is to write a stricter policy and collect acknowledgments. Lund recommends that too, and he's right that it's necessary. But a policy describes what people should do. It doesn't give a broker any view of what they actually did, and it doesn't answer the question an agent has at 9 p.m. with a contract open: where am I allowed to put this?

What agents use AI for tells a broker where the exposure sits

The NAR report also shows what agents use AI for, and the pattern matters more to a supervising broker than the headline adoption rate. Among agents who use AI, the top uses are the ones that leave the building.

The first three are public-facing, which is the territory where a regulator or an unhappy client can see the output. The last, document review, is the one that touches confidential transaction data, and it's the least-used of the four. It is also the use with the most room to gain from automation, since reading and tracking contract packages is repetitive work that sits squarely in the time agents say they are adopting technology to save. Lund's draft policy language specifically calls for prohibiting entry of "confidential client information, transaction documents, or personally identifiable information into unapproved tools." The less sanctioned a place there is for that work, the more likely an agent is to improvise one.

California adds a concrete example on the marketing side. A HousingWire report on a Coraly study published June 24, 2026 found 10.8% of roughly 40,000 sampled primary listing images showed indicators of digital manipulation, and more than 90% of the altered ones showed no visible disclosure. The same report notes that California Assembly Bill 723, in effect since January 1, requires licensed brokers and salespersons to conspicuously disclose digitally altered images. An agent editing listing photos in good faith can create a disclosure problem the broker never knew existed.

Bans don't work because agents are using AI to get time back

A real estate brokerage AI policy that amounts to a ban on agent AI tools fails for a reason the NAR data states plainly: 81% of agents said saving time is their primary goal in adopting new technology, up from 66% a year earlier, per the same NAR release. An agent who is told to stop using the tool that gives back an evening doesn't stop needing the evening. The usage goes quieter.

That reframes the question for an owner. The goal isn't fewer AI tools in agents' hands. It's a place where the useful work can happen in view of the brokerage, tied to the file it belongs to, with a record of what was produced. A shop that offers that gives agents a reason to follow the policy, and gives the broker something to point to besides the policy.

Why a 20-agent brokerage can see its whole perimeter

The independent has an advantage here that rarely gets named. Lund observes that "the larger the brokerage, the greater the risk," and notes that one incident is excusable while dozens or thousands are far worse. A principal broker with a few dozen agents can realistically know every tool in use, talk to every agent about it, and put one sanctioned system under the whole office. A very large organization has a much harder time having that conversation with each licensee.

That's a different edge from the one usually argued for small shops, which is speed. Visibility is about knowing. It's closer to the argument we made about AI compliance for independent brokerages, where consistent records do the work headcount can't, and it applies to AI itself, not just the paperwork AI touches.

How Loqol helps with sanctioned AI

Charlie AI, the assistant inside Loqol (loqol.ai), an AI and automation platform built for licensed brokerages, automates drafting, assembling, tracking, scheduling, compliance review of executed contracts and disclosure packages, analysis and number-crunching (comps, days-on-market, price history), vendor organization (inspectors, photographers, appraisers, escrow, title), estimating (repair credits, closing dates, timelines), and project management for agents, brokers, TCs, marketing, and admin alike. See the Charlie AI section for how that fits together.

For the supervision problem above, the relevant feature is placement. When the drafting, document assembly, and compliance review that agents want from AI happen inside the platform where the transaction file already lives, the broker has one sanctioned place to point agents to, and the work stays attached to the file it belongs to. Agents get hours back for clients, and the principal broker gets a clearer view of what automation touched, which is a better starting point than a signed acknowledgment alone. The same logic covers the document-review use that agents report least in the table above, which our piece on the paperwork gap in agent AI use looks at from the agent's side.

The conversation worth having with the broker

The real question behind a real estate brokerage AI policy is no longer "do we have an AI policy" but "where does agent AI work happen, and can we see it." Agents are already answering the first half on their own: they use AI to save time, mostly on content that leaves the office. Lund's instruction, "Treat AI as you would an unlicensed assistant," is a sound principle, and an unlicensed assistant works best when they're given a desk in the office, not a login on a personal phone.

For an agent who wants to push this internally, the argument is simple and favors everyone. A sanctioned AI and automation platform protects the broker's license, and it also protects the agent's own time, because agents get the benefit without the guesswork about what's allowed.