AI voice agent for real estate: what it should handle, what it costs and the rules to follow

The short answer

An AI voice agent for real estate is software that answers and makes phone calls in a natural voice: it picks up sign calls and listing inquiries, asks the qualifying questions your agents would ask, books showings and consults on the right calendar and logs every call in your CRM. It pays off when good leads ring through to voicemail because agents are at showings, at closings or asleep.

Use it first on inbound calls, where the caller chose to call you. Treat outbound AI calling with much more care: since February 2024 the FCC has treated AI-generated voices as artificial voices under the TCPA, so cold calling expireds, FSBOs or purchased lists with an AI voice carries real legal exposure. Inbound first, outbound only with documented consent, is the safe order.

What an AI voice agent for real estate should handle

Before you look at vendors, list the calls your team gets and decide what the AI does with each. This decides whether the system books appointments or just creates callbacks.

Call type	What the AI should do	Hand to a person when
Sign call or listing line	Answer from the approved listing sheet (price, beds, baths, square footage, HOA, showing instructions), ask if the caller has an agent, book a showing or a call with the listing agent	Caller wants to make an offer or negotiate
Buyer portal or website lead	Confirm area, price range, timeline and whether they are pre-approved or paying cash, book a buyer consult	Caller wants to tour today or is ready to write
Showing request	Check the showing calendar, offer windows, book and send a text confirmation	Occupied homes with special access rules, if the listing agent requires it
Seller "what is my home worth"	Collect address, condition and timing, book a listing appointment. Never quote a value	Seller wants to list within weeks
Rental inquiry	Answer availability, rent, deposit and pet policy from approved facts, book a viewing, send the application link	Questions about screening decisions or accommodations
Existing client or transaction	Take a detailed message, notify the agent or transaction coordinator	Anything about inspection results, appraisal, contract terms or closing funds
Lenders, vendors, sales calls	Take a message or politely end the call	Rarely

The rule of thumb: let the AI qualify and book, and keep negotiation, contract questions and anything involving money movement with licensed people. Wire instructions in particular should never be discussed by an AI agent; tell callers to confirm them only by phone with a known number. For the wider set of brokerage workflows around calls, such as transaction checklists and past-client nurture, see our real estate automation page.

Inbound vs outbound: where real estate AI calling gets risky

This is the section most vendor pages skip, and it is the one that can cost the most. It is practical guidance, not legal advice; have your broker of record and an attorney approve scripts and consent language.

• AI voices are artificial voices. On 8 February 2024 the FCC issued a declaratory ruling that AI technologies which resemble human voices or generate call content with a prerecorded voice fall within the TCPA's "artificial or prerecorded voice" rules. Kelley Drye's summary of the ruling notes the FCC said the TCPA has no carve-out for technology that claims to be the equivalent of a live agent.
• Marketing calls need written consent. Under that ruling, AI-voice calls that include advertising or telemarketing need the called party's prior express written consent, must identify the business responsible and must offer an opt-out. Calling someone off a purchased list, an expired listing or a FSBO sign with an AI voice is the highest-risk use you can pick.
• The one-to-one consent rule was vacated. NAR's telemarketing page still says a new FCC rule requiring consent to one seller at a time took effect on 27 January 2025. In fact, on 24 January 2025 the Eleventh Circuit vacated that rule in Insurance Marketing Coalition v. FCC, holding that the FCC had exceeded its authority under the TCPA. The general consent requirements still apply, so read the consent wording on your portal and website forms closely.
• Do Not Call still applies. NAR advises agents to check numbers against the National Do Not Call Registry before cold calling or texting. That applies to AI dialing too.
• Inbound is different. When a buyer calls your listing line, the AI is answering a call they placed. Recording and disclosure rules still apply: California and several other states require every party to consent to a recording, so open with "this call may be recorded" and say the caller is speaking with an automated assistant.
• Callbacks to people who asked. Calling back a lead who submitted a form and asked to be contacted is a common, lower-risk outbound use, as long as the form's consent language covers automated or AI calls from your business. Keep the consent record with the lead in your CRM.

Fair Housing guardrails for AI scripts

An AI agent says whatever its instructions allow, thousands of times. The Fair Housing Act prohibits discrimination based on race, color, religion, sex, national origin, familial status and disability, and steering buyers toward or away from areas is a known risk in agent conversations. Build these rules into the script before launch:

• Describe the property, not the people. The agent can state facts about the home, the lot and the listing. It should not describe who lives in a neighborhood or whether an area is "good for" a type of family.
• Deflect school and safety questions to sources. Point callers to official school district and public crime data sites rather than giving opinions.
• Ask every caller the same qualifying questions. Budget, timeline, financing and area preference, in the same order, regardless of accent, name or anything else the caller reveals.
• Handle accommodation requests carefully. Requests related to disability, such as service animals in a rental, should be logged and passed to a person, not answered by the AI.
• Review transcripts. Spot-check a sample every week, especially in the first month, and fix any answer that drifts.

Built-in, standalone or custom: your three options

You can get real estate AI calling in three ways. Some real estate CRMs and dialers have added AI calling or texting features (check what your plan includes), standalone AI front desk products sell monthly plans, and agencies build custom voice agents on platforms such as Retell AI and Twilio connected to your CRM.

Question	CRM or dialer add-on	Standalone AI front desk	Custom voice agent
Setup effort	Lowest, if you already use that CRM	Low. Quick to switch on	Highest. Scoped and built for your team
CRM logging	Native	Varies; some only email you a summary	Writes people, notes and tasks through the CRM's API
Listing knowledge	Depends on the vendor	Usually a manual knowledge base you keep current	Can read your IDX or listing feed so answers stay current
Routing rules	The CRM's own rules	Basic	Your rules: round robin, ponds, listing agent first, language, price band
Compliance controls	Vendor settings	Vendor settings	You decide disclosure, recording notices, consent checks and DNC scrubbing
Who owns the number and data	The platform	The vendor	You, in your own accounts
Best for	Teams happy with their CRM's defaults	Solo agents with simple call patterns	Teams and brokerages with several lead sources, listing lines and routing rules

If your CRM's add-on books the way your team works, start there. A custom build earns its place when you need several systems tied together, listing-aware answers or reporting across lead sources. The guide to choosing an AI automation agency lists what to ask any builder.

What an AI voice agent for real estate costs

Costs fall into three buckets: usage (minutes and phone numbers), any platform subscription, and the one-time work to connect the agent to your CRM, calendars and listing data.

Usage, from a published price list. Retell AI, a usage-based voice agent platform, lists pay-as-you-go pricing of $0.07 to $0.31 per minute depending on the language model and voice chosen. That is built from $0.055 per minute for its voice infrastructure, about $0.015 per minute for standard US telephony and model and voice charges on top. It also lists $2 a month per Retell phone number and 20 concurrent calls included (Retell pricing page, checked October 2026). Retell bills for the full call duration, including silence.

A worked example. Say a 6-agent team routes after-hours calls, listing-line overflow and requested portal callbacks to the AI: 250 calls a month at an average of 4 minutes, so 1,000 minutes.

Item	Monthly cost	Notes
Usage at the low published rate ($0.07/min)	$70	Lighter model and standard voice
Usage at the high published rate ($0.31/min)	$310	Premium voice and a larger model
Five listing and team phone numbers	$10	At $2 per number per month
Build and integration	One-time	Fixed quote after scoping; depends on CRM, calendars and listing feed

The staffing comparison. The US Bureau of Labor Statistics puts the median hourly wage for receptionists at $18.27 (May 2025). An office staffed 50 hours a week leaves 118 hours a week uncovered, including evenings and weekends. Covering those hours with people at the median wage would cost about $2,156 a week before overtime and benefits, which is why most teams rely on voicemail, a live answering service or agents answering on their personal phones.

Run your own numbers. Pull last quarter's missed calls and after-hours leads from your phone system and CRM, estimate how many would have become appointments, and multiply by your appointment-to-closing rate and average commission. If one extra closing a year covers the cost, the case is simple. Our AI automation agency cost guide explains how build fees are usually priced.

Connecting the voice agent to Follow Up Boss, kvCORE and your calendars

The voice is the easy part. The value comes from what happens after the call, which depends on the integrations.

• CRM. Follow Up Boss publishes a REST API for people, notes and related records, and asks developers who build for its customers to register their system. A custom agent can create or match the contact, attach the transcript and summary as a note, add tags such as "pre-approved" or "listing 90 days" and create a task for the assigned agent. kvCORE, Lofty and Sierra Interactive each have their own integration options; confirm what your plan allows before you scope a build.
• Calendars and showings. The agent needs read and write access to the calendars it books on, whether Google, Outlook or a showing tool such as ShowingTime. Decide which agent's calendar each call type books on, and set buffers for drive time.
• Listing data. To answer listing questions accurately, the agent should read from your IDX feed or a listing sheet that updates when status or price changes. Stale data, such as describing a home that went pending yesterday, is the most common complaint about AI listing lines.
• Lead routing. Match your existing rules: listing agent first for sign calls, round robin for portal leads, Spanish-speaking agents for Spanish callers, ponds for unclaimed leads.
• Workflow layer. Tools such as n8n handle the glue: post-call text confirmations, Slack or SMS alerts for hot leads and a weekly report into Looker Studio.

If most of your problem is follow-up after the first call rather than answering it, CRM automation or lead generation automation may matter more than the voice agent itself.

How to set up an AI voice agent for your real estate team, step by step

1. Pull your call and lead data. Export 60 to 90 days of calls by hour, missed calls, voicemails and lead sources from your phone system and CRM. This shows where the AI should start, usually after hours and listing-line overflow.
2. Map call types and rules. Use the table above. Add your service areas, price bands, which agents take which leads, and what the AI may and may not say.
3. Fix consent and disclosure first. Review form consent language, recording notices and AI disclosure with your broker and attorney before any outbound use.
4. Connect the CRM, calendars and listing feed. Use a service account rather than one agent's personal login.
5. Write the scripts. Greeting with recording and AI disclosure, qualifying questions per call type, Fair Housing guardrails, hand-off and hot-lead rules.
6. Test on real calls. Replay 30 to 50 real past calls, including callers with heavy background noise, Spanish speakers if you serve them, and callers who ask about schools or neighborhoods. Fix every wrong answer.
7. Go live narrow. Start with after-hours calls on one or two listing lines, and have an agent or ISA review every booking each morning for the first two weeks.
8. Review weekly, then monthly. Track answered, booked, transferred and abandoned calls, and appointments kept, by lead source.

A CRM add-on can be switched on in a day, though testing still applies. A custom voice agent with CRM logging, calendar booking and listing-aware answers typically takes 4 to 8 weeks, delivered in stages.

Common mistakes, and when not to buy

• Starting with outbound cold calling. It is the use with the most consent risk and the least goodwill. Prove the agent on inbound first.
• Letting the AI give home values or advice. Valuations, offer strategy and contract questions belong with licensed agents.
• Stale listing answers. If the agent cannot see status changes, callers get wrong information and lose trust.
• No owner for unbooked calls. Every call that ends without an appointment should land on someone's task list the next morning.
• Measuring calls answered instead of appointments kept. The second number is the one that leads to closings.

Skip it or wait if: your team already answers nearly every call live with an ISA or answering service you are happy with; you get few calls and most leads come by text or chat; or your CRM data is messy enough that the AI would create duplicates and misroute leads. Clean the CRM first. If you mainly manage rentals, our AI receptionist for property management guide covers maintenance and leasing calls in more depth.

Getting an AI voice agent built for your real estate team

If your CRM's built-in AI calling covers your call types and routing, use it. Bring in an agency when you run several listing lines and lead sources, need listing-aware answers, want your own consent and disclosure controls, or want call outcomes in one report.

We build AI voice agents and AI receptionists for real estate teams and brokerages, connected to Follow Up Boss, kvCORE, Lofty and the calendars you already use. Every project starts with a free review of your call and lead data, gets a fixed price, ships in stages and runs on accounts you own, including the phone numbers, scripts and code.

To get started, book a free review and tell us which CRM you use and roughly how many calls go to voicemail each week.

FAQ

Frequently asked questions