Although it was first launched as a generative technology to be used as a tool, AI is now “agentic,” meaning it can function as an operator. Translation: AI can now be configured to do everything from drafting and sending messages to updating records in a CRM to scheduling follow-ups and triggering workflows—without human intervention. Scary.
If you’re a broker, what it means for you and your agents is improved productivity and consistency—as well as the ability to build out customized tools for your brokerage.
But there are serious risks for those who aren’t prepared.
David Conroy, the National Association of REALTORS® chief data officer, laid out the risks and made several recommendations to brokers attending the Idea Exchange Council for Brokers Forum during the 2026 REALTORS® Legislative Meetings in Washington, D.C.
Risks include:
- Data provenance and MLS compliance issues. Because AI can trigger a workflow without nuance, it may act on stale, incomplete or non-compliant data. And since MLS rules aren’t machine readable, AI could lead to risk of accidental violations.
- Fair housing exposure. Automated messaging or recommendations could inadvertently create a discriminatory pattern. It’s important to remember that even unintentional bias is a fair housing liability.
- Branding and advertising risks. AI has been known to hallucinate or generate information that sounds correct but is not. With branding, it can generate unapproved claims, remarks or marketing. It might also miss disclaimers or use the wrong logo.
Taking on responsibilities that require a license. Brokers know which activities— providing formalized professional advice, creating listings and so on—require a license. But if these things are built into a workflow that’s been handed off to an agentic AI tool, that tool could move forward without waiting for a licensed practitioner to take over at the appropriate time.
“Things can get really dicey when your AI tool is acting in similar ways as a licensed activity,” Conroy said.
These are only some of the risks agentic AI poses, added Conroy, emphasizing the need for a formal policy. “Agents are already using this technology. They are already experimenting with it,” he said. It’s up to brokers to take the reins and implement a policy that protects clients and the brokerage.
Conroy made the following recommendations:
- Evaluate and choose the tools your brokerage will use. Because your agents are already experimenting with various AI tools, it can create a sense of disconnection and make it hard for you to oversee their use of AI. Ideally, Conroy suggested that, ideally, brokers should vet tools and choose a suite of approved options. If an agent is comfortable with a tool that’s not on your vetted list, however, stay flexible. Find out what tools agents are using and how. Then, figure out how to migrate them over to your approved tools, or allow them to use that tool in ways that fit into your brokerage’s AI policy parameters.
- Implement mandatory human review. Nothing produced with AI should go out without human oversight. Someone at the brokerage, whether its staff or a designated leader, should have eyes on everything. Designing an approval process can help with this.
- Come up with a risk management enforcement plan. Conroy noted that NAR uses a tiered approach, with consequences that align with the severity of the infraction. He encouraged brokers to come up with an enforcement plan that meets the needs of their business.
- Get a policy out sooner rather than later. It’s more important to get a policy into the hands of agents than to have it completed fleshed out. The tools keep changing—and at a breakneck pace—so the policy will have to evolve anyway. Conroy advised brokers to get an accessible and reasonable policy together and to start the conversation with agents soon. A more comprehensive policy could be architected over time.
NAR’s AI policy template for brokers is completely customizable and has both short-form and long-form options. It’s a good place to get started.
