Your Next Real Estate Assistant Will Be a Phone Call Away

A realtor gets back into the car after a showing. The buyers liked the house, hated the road, and mentioned that their mortgage approval expires next month. They want to see two more properties on Saturday. One of them asked about basement apartments.

There are at least five pieces of work in that conversation.

Usually, those tasks travel around with the realtor for the rest of the day. They become a voice memo, a mental note, an unanswered text, or another reason to open the laptop after dinner.

Now imagine making a call to your AI assistant before leaving the driveway.

“Update the buyers’ notes. They want a quieter street, and we need to confirm their financing timeline. Find a few options for Saturday and draft a follow-up with what we discussed.”

The assistant asks which Saturday time works. You answer. It starts the work. You remember another detail and add it without restarting the conversation.

That is the experience I think will change how realtors run their businesses: delegating to AI agents as naturally as calling someone at the office.

On September 10, OpenAI announced GPT-Live-1 in the API, opening its conversational voice technology to developers building their own applications. It can listen and speak simultaneously, handle interruptions, and delegate reasoning and actions to connected models and tools while the conversation continues. The release also supports telephony, making actual phone-based experiences possible. OpenAI’s announcement

Watch the original announcement on X

For real estate, the timing of the work matters almost as much as the work itself.

The best moment to capture a buyer’s reaction is immediately after the showing, while the details are fresh. The best moment to organize the next step is when you have just discussed it. Every hour between the conversation and the follow-up creates another opportunity to forget something, lose context, or get pulled into another deal.

Voice delegation could make that gap much smaller.

With the right connections and permissions, a realtor could speak to an assistant through the car’s hands-free system, give it a short briefing, and have work underway before reaching the next appointment. Reviewing documents and approving sensitive decisions would still happen once parked. The useful change is that the preparation no longer has to wait for a desk.

The following is an illustrative workflow, rather than a demonstration of a finished real estate product:

“Pull up the buyers from this afternoon. Add that the road noise was a dealbreaker.”

“Added. Should I use that preference when preparing the next shortlist?”

“Yes. And prepare a comparison of the three homes we saw. Use the listing information and my notes. Flag anything we still need to verify.”

“I’ll prepare that. What would you like to handle next?”

That last sentence is where this gets interesting. The realtor can continue the conversation while another agent prepares the comparison. One assistant could coordinate several tasks: updating records, gathering information, drafting communications, and checking a calendar.

OpenAI has already shown voice coordinating work across agents in ChatGPT Work and Codex. Its official demo below is from July 23, and the September API announcement links to it as an example of this approach. Watch the original on OpenAI’s YouTube channel.

I think real estate is particularly well suited to this because so much of the job happens away from a computer.

Agents gather useful information at kitchen tables, in condo lobbies, during inspections, and on the phone. Their software generally expects that information to arrive later, neatly organized into fields and forms.

A voice assistant could meet the realtor at the moment the information is available. The agent explains what happened in ordinary language, and the system turns that explanation into records and tasks.

That could change the economics of a small team.

A solo agent might be able to maintain more consistent follow-up without personally completing every administrative step. A team administrator could spend less time reconstructing incomplete notes and more time resolving exceptions. A team leader could ask for a briefing on outstanding work and direct the next steps in the same conversation.

Consider seller updates. After a weekend of showings, a realtor could ask an assistant to gather recorded feedback, summarize recurring objections, and draft an update for the seller. The realtor would review the evidence, add their interpretation, and decide what to recommend.

Or consider a listing appointment. Before leaving, the realtor could request a spoken briefing from the client record: what the seller said last time, what information remains outstanding, and which questions need an answer. The value comes from arriving prepared without having to assemble the briefing manually.

Even content creation becomes easier to fit into the day. An agent could explain an observation after a showing and ask an assistant to turn it into a draft post, removing identifying client details. A useful local insight has a better chance of becoming something publishable while it is still fresh.

There is still work between this announcement and a dependable real estate assistant.

The API provides the voice capability. Someone still has to connect the CRM, calendar, email, documents, and authorized property data. The system needs to identify the correct client, preserve instructions, and report what actually happened. A showing request must remain a request until the appointment is confirmed. An email draft must remain a draft until it is sent.

Those details will determine whether agents trust it enough to use it every day.

My expectation is that the first substantial gains will come from repeatable work with clear outcomes: capturing showing notes, preparing follow-ups, assembling appointment briefings, and tracking outstanding tasks. Each successful handoff makes the next one easier to trust.

For brokerages, this creates an opportunity to improve the service behind every agent. They can build shared workflows so that a new agent and an experienced producer both have help turning conversations into completed next steps. Better systems could reduce the cost of delivering that service and give people more time for clients.

The adoption hurdle also becomes much smaller when the starting point is a conversation. A realtor who can explain a task to an assistant already understands the basic interaction. The technology has to do the harder work of translating that request into a reliable result.

That is why this announcement matters to me. It brings us closer to a working day where leaving a showing triggers the follow-up, where a conversation produces a useful record, and where delegating work takes a brief call.

The realtor gets to the next appointment. The work from the last one is already moving.