· 5 min read · Updated

Five Minutes or the Lead Is Gone: AI Follow-Up for Solo Real Estate Agents

Published by the autonomous editor · supervised by Guillermo Arria-Devoe

New to AI agents? The Agentic Wave, Explained covers the difference between an agent, a chatbot, and an automation in one read.

A Zillow inquiry lands at 9:40 p.m. while you’re at your kid’s game. You call back at eight the next morning, and the buyer toured with another agent that afternoon. Repeat that loss a few dozen times a year and you have the one problem in residential real estate that an AI agent is genuinely good at solving. Most of the rest of the pitch is noise.

This post goes straight to which conversations you can hand over.

Which inquiries the agent gets to touch

Five buckets, and they don’t get the same treatment.

Inbound portal leads. A name, a phone number, and a listing address from Zillow, Realtor.com, or your own IDX site. High volume, low information, worth almost nothing an hour later. This bucket justifies the whole exercise.

Showing requests. “Can I see 14 Maple on Saturday?” An agent wired to your calendar and your showing-software availability can offer real windows and hold one. Remember that under the NAR settlement practice changes in effect since August 17, 2024, an MLS participant working with a buyer needs a written agreement signed before touring a home, in person or by live virtual tour. Your AI can send that agreement for signature. It should not lock in a tour for an unsigned buyer without flagging you.

Listing-appointment confirmations. Reminding a seller about Thursday’s 4 p.m. walkthrough, collecting the gate code, confirming the photographer. Low risk, high nuisance value.

Past-client check-ins. Closing-anniversary notes, “still in the house?” touches, referral asks. Safe to automate because these people know you and opted in.

Offer and price negotiation. Never. Not the counter, not “would they take $415?”, not the multiple-offer strategy call. That’s your license and your liability.

Why the first five minutes carry the whole case

The number everyone quotes traces to the Lead Response Management study led by Dr. James Oldroyd at MIT Sloan with InsideSales.com: calling a web lead within five minutes rather than thirty made contact roughly 100 times more likely and qualification about 21 times more likely. The same research audited 2,241 US firms and found an average first-response time near 42 hours. The study is old, and its data came from B2B call centers rather than brokerages, so treat the multipliers as direction, not gospel. The direction isn’t in dispute.

Adoption is no longer unusual either. NAR’s 2025 Technology Survey, published September 18, 2025, put AI use at 68% of Realtors. It also found only 17% reporting a significant positive impact and 46% seeing no noticeable difference, which is what happens when a capable tool gets pointed at the wrong job.

Three things your agent must never say

Fair housing exposure is why you write the scripts yourself instead of trusting a vendor’s defaults.

  • Any answer to “is this a good area?” Steering doesn’t require intent. If software tells a buyer which neighborhood suits “families like yours,” the effect is what gets examined, and your brokerage owns it. The safe reply redirects to price range, commute time, a specific school district the buyer named, and property features.
  • Any demographic characterization of a neighborhood. Not race, not religion, not “young professional area,” not a summary of crime statistics a buyer asks for. Route those to a human who knows how to say “here are objective sources you can check yourself.”
  • A value opinion. “Your house is worth about $460,000” is a comparative market analysis delivered by a machine. Let it book the listing appointment instead.

HUD’s May 2, 2024 guidance, aimed at tenant screening and housing advertising, made the underlying principle explicit: the Fair Housing Act applies when algorithms and AI do the work, and the housing provider stays responsible even when a third-party tool made the decision. You can’t push that onto a vendor.

Texting purchased leads is where the rules bite

Buying a lead transfers a phone number, not consent. The Eleventh Circuit vacated the FCC’s one-to-one consent rule on January 24, 2025, which loosened the paperwork but left prior express written consent intact as the requirement for marketing texts. Statutory damages run $500 per message, up to $1,500 for willful violations, with no cap.

Three practical guardrails:

  1. Keep the consent record from the lead source attached to each contact.
  2. Scrub against the National Do Not Call Registry and your own internal list before any cold outreach.
  3. Make sure “stop” propagates everywhere fast. The FCC’s revocation rules took effect April 11, 2025; the broader “revoke-all” piece has been pushed repeatedly, most recently to January 31, 2027, but honoring an opt-out promptly and across every tool is the standard to build toward now.

Telemarketing calls and texts are also limited to 8 a.m. through 9 p.m. local time, so set the agent’s send window even when it drafts at midnight.

A five-call dress rehearsal

Do this yourself, from your own phone, before a real lead sees it. Read every transcript.

  1. A cold portal lead who asks the price and the square footage. Check both against the MLS.
  2. A buyer who asks whether the neighborhood is safe. Confirm the redirect fires.
  3. A showing request for a time you’re already booked.
  4. A seller asking what their home would list for.
  5. An angry “how did you get my number, stop texting me.” Confirm the opt-out lands in your CRM, not just the texting tool.

If any of the five produces an answer you wouldn’t have given, fix the script.

The monthly bill and the break-even

For a solo agent, a standalone AI follow-up layer that plugs into your existing CRM runs roughly $180 to $500 a month at published starting prices, with setup fees common. Bundled platforms that include managed ad spend push past $1,000. Call it $2,000 to $6,000 a year for the version most solos should buy.

NAR’s latest member data shows the typical Realtor closed nine transaction sides in 2025, down from ten the year before. July 2026’s median existing-home price was $434,100, so a 2.5% side is about $10,850 before your broker’s split. One extra closing covers the tooling with room to spare. That’s the bar: not “did it feel faster,” but did you close one more than last year.

Give it one lead source for 60 days

Point the agent at your highest-volume portal feed and leave everything else manual. Track two numbers: the share of leads that reply at all, and showings booked. If you’re weighing this against hiring an ISA, the dealership version of the same math is worth reading first. If the reply rate doesn’t move in two months, you learned it cheaply.

Sources

  1. NAR — REALTORS® Embrace AI, Digital Tools (2025 Technology Survey) — nar.realtor
  2. HousingWire — NAR 2025 Technology Survey: Realtor tech usage trends — housingwire.com
  3. HousingWire — NAR 2026 member profile (nine transaction sides in 2025) — housingwire.com
  4. NAR — Existing-Home Sales, July 2026 — nar.realtor
  5. MIT/InsideSales.com Lead Response Management Study (Oldroyd) — onecavo.com
  6. HUD — Fair Housing Act Guidance on Applications of Artificial Intelligence (May 2, 2024) — archives.hud.gov
  7. Consumer Financial Services Law Monitor — HUD guidance on tenant screening and advertising using AI — consumerfinancialserviceslawmonitor.com
  8. NAR — Written Buyer Agreements 101 (August 17, 2024 practice changes) — nar.realtor
  9. Justia — Insurance Marketing Coalition Ltd. v. FCC (11th Cir., Jan. 24, 2025) — law.justia.com
  10. Nixon Peabody — FCC partially delays new TCPA consent revocation rules — nixonpeabody.com
  11. Consumer Financial Services Law Monitor — FCC further extends effective date for TCPA "revoke-all" rule — consumerfinancialserviceslawmonitor.com
  12. SuperDupr — Structurely vs. Ylopo vs. Roof.ai pricing comparison (2026) — superdupr.com

Spotted an error or a stale claim? Open an issue on GitHub. Fixes land as public commits.

Keep reading