One Real Estate Team's AI Answered 900 Calls. When It Wasn't Sure, It Handed Off to a Person.

The National Association of Realtors published its 2026 Technology Report on 22 September. Nearly half of agents now use AI daily (23%) or weekly (25%). Only 12% say they don't use it and don't plan to. Among agents who do, the top uses are writing listing descriptions (75%), social media posts (56%) and emails and follow-ups (52%). And 81% say the reason they take on any new technology is to save time.

So the question has moved on from "should I try it?" to "which jobs should it take, and which should stay with me?" One small Ohio team answered that with a single rule.

LKW Home Team: 900 calls and a 95% rule

Lauren Lucas co-owns LKW Home Team in Columbus, Ohio. In a piece published on Real Estate News in March 2026, she described a voice AI agent that had answered more than 900 inbound calls and taken over 400 next-step actions from them. When a buyer calls, the assistant collects name, phone number, email, preferred neighbourhood and price range in under two minutes, and asks whether they are pre-approved.

The detail that matters is the limit. The system is trained on the team's procedures, compliance rules and market updates. If it is not at least 95% confident in an answer or a next step, it transfers the call to a person. Web enquiries follow the same pattern: the assistant creates the contact, tags it in the CRM and prompts a booking. Once the client books, an agent takes over, already knowing what the first call would have asked.

Two labels belong on this. The piece is bylined by Keller Williams, the brokerage brand the team is affiliated with, and the figures are the team's own, not an audit. The same article says the team shrank from seven agents to three between 2019 and 2023 while its gross commission income rose 35.8% going into 2024. It does not say how much of that came from AI, so neither will we.

Where agents' hours actually go

Writing is where most agents start, because it is low risk: a machine drafts, a person checks. The bigger time cost is often the inbox and the phone. In the same piece, Shoney Ivens, who leads about 300 agents at a Keller Williams office in Orem, Utah, estimated that the typical agent spends 12 to 15 hours a week in their inbox. He uses AI to draft emails and meeting recaps, and redacts sensitive details before he uploads a contract for a summary.

A rough order for a small team:

• Writing: listing copy, social posts, market summaries.
• Inbox and follow-up: drafted replies and reminders that a person approves. See a small studio that answered email in 14 seconds.
• Phones and intake: the LKW pattern. Structured questions, a CRM entry, a booking, and a hard hand-off rule.
• Documents: summaries of contracts and disclosures, personal details removed, checked by a person.

Read the vendor case study like a contract

Search for AI tools for real estate agents and most of what you find is vendor case studies. Read them closely. Realty AI's customer page for Mary Wylde, a RE/MAX broker in Guelph, Ontario, dated May 2026, says its assistant handled 1,896 website conversations and that six listings worth over $4.8 million closed from them. It also claims a "42X ROI" on commission income. The same page gives the new leads as 434 in its headline figures and 343 in its results list, and the hours saved as 320 in one place and 382 in a heading.

That doesn't mean the tool didn't help. It means a marketing page is not a measurement. Ask for the raw export: leads, appointments and closings, with dates, against the year before.

Where judgment stays with a person

Some jobs look automatable and aren't. Listing photos are the clearest case. Since 1 January 2026, California law has required agents who use photo editing software or AI to add, remove or change elements of a listing image to say so next to the image and link to the unaltered original. Virtual staging that hides a cracked wall is now a regulatory problem, not a clever shortcut.

The same goes for pricing advice, contract terms and anything touching fair housing. A machine can draft; a licensed person decides.

Someone also has to own the assistant. A voice agent quoting last month's price or a sold listing loses you the caller. Name one person who reads a sample of transcripts each week and keeps scripts, listings and prices current.

How a small team should pick its first AI tools

1. Count the hours by job for two weeks: writing, inbox, phones, paperwork. Agents in NAR's survey named the learning curve (63%) and cost (59%) as their biggest obstacles, so start where the hours are.
2. Pick one job and one tool that connects to the CRM you already use.
3. Write the hand-off rule before you switch it on: what the AI may do alone, what it drafts for approval, and what goes straight to a person.
4. Name the owner who reviews transcripts and keeps the facts current.
5. Measure against last year: appointments and closings from that channel, not "conversations handled".

This is how we think about AI in any small firm. Let it take the routine work, such as intake questions, first drafts and reminders, so people's attention stays on judgment: pricing, negotiation and the client who needs advice. If phones are your bottleneck, our guide to the best AI receptionist and answering service for small business covers the options, and our write-up of what AI customer service agents actually finish explains why the hand-off decides the return. If you want help choosing the job, connecting it to your CRM and setting the rules, that is the work of our automation service.

Frequently asked questions

What are the best AI tools for real estate agents?

Start from the job, not the brand. In NAR's 2026 survey, agents who use AI mostly use it for listing descriptions (75%), social posts (56%) and emails (52%). Phone answering, intake and document summaries come next. The best tool fits the job that costs you the most hours and plugs into your CRM.

Can an AI assistant answer calls for a real estate team?

Yes, for structured calls. LKW Home Team in Columbus, Ohio says its voice agent has answered more than 900 calls. Copy its rule: below 95% confident, the call goes to a person.

Do I have to disclose AI-edited listing photos?

In California, yes. Since 1 January 2026, Business and Professions Code section 10140.8 requires agents to label digitally altered listing images, including AI edits that add, remove or change elements, and to link to the original. Check your own state and MLS rules before you virtually stage a room.

How much do AI tools for real estate agents cost?

In NAR's 2026 survey, 36% of agents spend $50 to $250 a month on all their technology. Price the hours a tool saves on one job before adding another subscription.

Tell Ada which job eats most of your week: listings, inbox, phones or paperwork. She will suggest what an AI tool could take, what should stay with your team, and put it in your free report. Talk to Ada and request your free report.

Sources: National Association of Realtors, "Realtors Adopt Technology to Save Time and Improve the Client Experience, NAR Report Finds", 22 September 2026, and the 2026 Realtors Technology Report key findings. "Do more while doing less: Keller Williams-affiliated agents share top AI plays", Real Estate News, 9 March 2026 (bylined Keller Williams; figures are the agents' own). Realty AI customer page "Mary Wylde", 29 May 2026 (vendor claims about its own product). California Business and Professions Code section 10140.8, added by AB 723 (2025), effective 1 January 2026.