9:42pm
Nobody Leaves a Voicemail Anymore. They Call the Next Agent.
A missed call leaves no name, no message and no record that anybody wanted you. Here is what an AI voice agent actually does when the phone rings at 9:42 on a Sunday, the one thing that decides whether it works, and the disclosure rules nobody selling one mentions.
The phone rang at 9:42 on a Sunday evening. Somebody had been on your listing for eleven minutes, got as far as the taxes, and wanted a person.
You were at dinner. Four rings, then your voicemail, then the small silence where somebody decides what to do next.
They did not leave a message. Almost nobody does. They went back to the results page and called the next number on it.
That is the whole problem, and it is not really a lost lead. It is a lost conversation, which is the only place a lead has ever turned into anything.
In short
- A missed call is not a message waiting for you in the morning. Most callers never leave one, so it arrives as nothing at all: no name, no number worth calling back, no record that anybody wanted you.
- Speed decides who gets the conversation. Across 1.25 million leads, firms that made contact within the hour were nearly seven times likelier to get a meaningful conversation with a decision maker than firms that waited one more hour.
- An AI voice agent answers every call, qualifies inside a normal conversation, books from your real calendar and writes the outcome to your CRM. It does not close, and it says out loud that it is an assistant.
Watch it work
A phone ringing at 9:42, and the same call taken properly.
Narrated, 45 seconds. The voice reading it is a licensed clone of my own, which on this page seemed worth saying out loud rather than leaving you to wonder. The call is staged for the film and every line in it is something this page already says the agent does. The seven-times figure is the study cited below. For the version nobody staged, the service page has the whole build.
What a missed call actually costs
We could not find a published study of how fast real estate agents answer their phones. What has been published is the next best thing, and it is worth reading properly rather than quoting.
In 2011, Harvard Business Review reported an audit of 2,241 US companies. The researchers submitted a test inquiry through each company's own website and timed how long the reply took.
The evidence
How long 2,241 companies took to answer an inquiry from their own website
Share of the 2,241 companies audited. Source: Harvard Business Review, The Short Life of Online Sales Leads (2011), reporting the audit by James Oldroyd, Kristina McElheran and David Elkington.
Cross-industry research from 2011, timing replies to inquiries submitted through a website rather than to phone calls. These percentages are not a real estate benchmark and should not be quoted as one. The shape is the part that carries over.
Twenty three percent of them never replied at all. Among the ones that answered inside the thirty day window, the average took 42 hours.
Two honest notes before that number gets used for anything. It is cross-industry research from 2011, not a real estate study from this year, and the inquiries were web forms rather than phone calls. What carries over is not the percentage. It is the shape: a large group answers quickly, a large group never answers, and the average sits in days rather than minutes.
The same researchers ran a second study, this one across 1.25 million sales leads at 29 business-to-consumer and 13 business-to-business companies. Firms that tried to make contact within an hour were nearly seven times likelier to qualify a lead than firms that waited one more hour, and more than sixty times likelier than firms that waited a day.
The definition is the part everyone skips over. They counted a lead as qualified when it produced "a meaningful conversation with a key decision maker". Not an email opened. Not a form completed. A conversation, with the person who decides. That is exactly what your 9:42 call was, seen from the other end, which is why this research applies more directly to a ringing phone than to most of the things it gets quoted about.
Why nobody leaves the voicemail
Ask anyone what they do when a business does not pick up and you get the same answer, usually with a slightly embarrassed shrug. They hang up and try somebody else.
It is not rudeness and it is not impatience. Leaving a voicemail is a commitment to wait, made by a person who has four more numbers on the screen in front of them and no particular reason to prefer yours. It also hands the next move to you, and they have no way of knowing whether you check.
So what you receive is not "a lead could not reach me". It is nothing at all. There is no artifact. It shows up as a missed call, which looks identical to a wrong number, and by Tuesday it is gone.
That is the specific reason a phone gap is worse than an inbox gap. An unanswered email is still sitting there in the morning with somebody's name on it, and you can still be the one who replies. An unanswered call left no name, no number that is worth ringing because they are already talking to somebody else, and no record that anything happened.
In your numbers
What do the calls you never took add up to?
Evenings, weekends, showings, the school run. Count the ones that rang out.
Wrong numbers, vendors and the title company are not leads.
When do you usually call them back?
Going to whoever picked up
$27,000a year
- Calls that rang outyour 25
- 25a month
- That were a real inquiryyour 60%
- 15a month
- You never get back into a conversation withhalf of them
- 7.5a month
- Over a year12 months
- 90a year
- That would have become a closinga 5% close rate
- 4.5a year
- In commissionyour $6,000
- $27,000a year
Two things worth knowing about this arithmetic. The first is that it is far gentler than the research above would justify: the 1.25 million lead study found firms contacting within an hour were nearly seven times likelier to get a meaningful conversation with a decision maker than firms that waited one more hour, and more than sixty times likelier than firms that waited a day. Turning odds like that into a slider would produce a number nobody should believe, so the callback curve here is a plain judgement and yours to disagree with. The second is that the 5% close rate is the same conservative one the website chat piece uses, even though somebody who dialled you is warmer than somebody who filled in a form. Neither post gets to tune its own number. This counts inbound calls only, and the contacts already sitting cold in your CRM are a different population in a different article.
What an AI voice agent actually does
Stripped of the marketing, it is a voice on the phone that can hold a real conversation, wired to the systems that make its answers true. Ours runs on Vapi.
What it actually does
Four moves.
It answers every call, including the ones at nine on a Sunday.
Inbound calls route to the agent whenever you are unavailable, or always, depending on how you set it up. It greets the caller, answers what it can from your listings, and never sends anybody to voicemail.
It calls new leads back in seconds, not in the morning.
The moment an inquiry lands from your site, a portal or a campaign, the agent dials. It is calling while the person is still on the page, which is the only window that has ever really been open.
It qualifies inside a normal conversation.
Budget, area, timeline, whether they are pre-approved, whether they have a house to sell. Not a script read at somebody. The questions you would ask, in the order a conversation actually goes.
It books, and then it writes it down.
Real availability from your calendar, offered on the call and confirmed on the call. Afterwards the transcript, the qualification and the outcome go into your CRM, so the appointment you walk into is one you already understand.
The fourth one is the one people underrate. When you ring back, you are not opening with "how can I help". You already know they got as far as the taxes, that they asked twice about the septic, and that they used the word soon. On the phone that is worth more than it is anywhere else, because you can hear how somebody answers a question they have already answered once, and whether the second version is the same.
The same agent, pointed the other way
Everything above is about the call you did not take. The bigger half of this is the call nobody made.
Most leads never phone you at all. They fill in a form at nine at night, or they arrive from a portal or an ad, and then they sit in a list waiting for somebody to notice them. The standard answer to that is a follow-up task, which gets done tomorrow, or on Thursday, or never.
The second of the four moves above is that other direction, and it earns more than the one line it gets there. It is not only the automatic call either. You can point it at somebody and tell it to dial, and you can point it at a list.
The other direction
Four ways it calls out.
You can point it at one person and tell it to call.
It does not have to wait for a trigger. Somebody registers, you want them spoken to before they open the next tab, so you tell the agent to dial that contact and it does, with the same qualifying conversation and the same booking at the end of it.
It calls again, on a schedule, without getting discouraged.
Somebody who did not pick up at two in the afternoon was driving, not uninterested. Human follow-up tends to die at the second attempt, because a third one starts to feel like pestering. The agent tries the evening instead, and the evening is often the one that connects.
It can work the list you already gave up on.
Everybody who enquired last year and went quiet is the cheapest inventory you own, and nobody calls it because calling it is miserable. The agent works through it, has real conversations, and hands back the few who are ready to move now.
It can tell a whole list at once, not one at a time.
A price drop, an open house, a listing that matches what forty people told you they were looking for. One caller works that list for a week. The agent works it in parallel, talks to whoever picks up, and books the ones who care.
The last of those is the one to be careful with, and it is worth saying plainly rather than selling. Calling a hundred people at once is the same act whether a person does it or software does, and the rules do not soften because the dialing got cheaper. You still need a reason to be calling each of them, you still have to honor do-not-call and the calling windows, and the agent still has to say what it is when somebody picks up. Volume is exactly where outbound stops being useful and starts being a complaint, so the list you call should be people who asked to hear from you about that thing. The legal section below is about outbound at least as much as it is about recording.
The one thing that decides whether it works
Text chat forgives a pause. A typing indicator is a promise that something is coming, and two seconds of it costs nothing.
A phone call has no typing indicator. There is only silence, and silence means one of three things to the person holding the handset: the line dropped, the other person is thinking, or nobody is there. Callers resolve that ambiguity in well under a second, and they resolve it badly.
How far under a second is worth seeing, because it is smaller than almost anybody guesses.
The size of the gap
How long people leave before answering a question, measured in ten languages
The grey track behind each bar is one full second. The outer two bars are the fastest and the slowest of the ten language averages; the inner two are the middle and the average of the whole dataset, which sit apart because the spread runs long. Ten languages on five continents, measured from video of ordinary unscripted conversation between two and six people. Yes or no questions were 67% of everything coded. Source: Stivers and ten co-authors, Universals and cultural variation in turn-taking in conversation, PNAS 2009.
Read what this does not say. Every conversation in it happened face to face and not one of them was a phone call, and the same paper found responses came earlier when the questioner was looking at the recipient in nine of its ten languages, reaching significance in five, which is exactly what a phone takes away. It measures people answering people, so it has nothing at all to say about what a caller will put up with from a machine. What it does establish is the size of the window. Every one of the ten languages had its commonest gap somewhere between nothing and two tenths of a second, and the authors note that speakers are sensitive to shifts of a hundred milliseconds. An agent that thinks for a full second has left more than twice the gap of the slowest conversational culture in the study, and the caller does not hear that as thinking.
So latency is not a specification on a phone agent. It is the product. The gap between a caller finishing their sentence and the agent starting its answer is the entire difference between "somebody at the office picked up" and "I got one of those robots". That is why this stack is built around sub-second response rather than around a longer feature list.
Two related things get forgotten, and between them they cost more calls than anything in a brochure.
People interrupt. Real phone conversations are full of overlap. Somebody starts answering before you have finished the question, you stop talking, and neither of you thinks about it afterwards. An agent that cannot be cut off mid-sentence talks over a human being instead, and being talked over by a machine is the precise moment people hang up.
Turns are short. On the phone a person says four words and waits. "Is it still available." An agent tuned to write paragraphs answers a four-word question with a speech, and the caller does exactly what you would do.
Silence on a phone is not a pause. Every second of it is the caller deciding whether anybody is actually there.
The teardown
Watch it take the 9:42 call.
The pattern, not a recording of a real client.
What the caller heard
- Caller9:42 pm
Hi. I am looking at the colonial you have listed. Is it still available?
- Agent9:42 pm
It is. I am the assistant for the office, and I record calls so nothing gets lost between now and the morning. What would you like to know about it?
- Caller9:43 pm
What are the taxes on it, and would the seller look at a contingent offer?
- Agent9:43 pm
The taxes I can read you straight off the listing. The contingent offer I am not going to guess at. Nobody knows until the listing agent is actually asked, and it is the kind of answer that decides an offer. Let me put you with a person in the morning.
- Caller9:44 pm
Morning is fine.
- Agent9:44 pm
Nine thirty is open. You are booked in, and I have sent the details to the number you are calling from.
What happened behind it
- 9:42 pm
Answered on the second ring
No hold, no voicemail, no callback form.
- 9:42 pm
Said what it was
An assistant, and the call is recorded. Before anything else.
- 9:43 pm
Answered from the record
Read off the listing rather than recalled.
- 9:43 pm
Declined to invent an answer
Flagged for a person instead of guessing at it.
- 9:44 pm
Booked from the live calendar
Only offered a time that was genuinely free.
- 9:45 pm
Written to the CRM
Transcript, qualification and outcome, before dinner ended.
What it does not do, and should not pretend to
This is the part vendors skip, so here it is plainly.
It does not close, and on the phone the reason is specific. Half of what a good agent hears in a first call is not in the words: the pause before an answer, the second voice in the background saying something you were not meant to catch, the cheerfulness that is covering something. A transcript records none of that, and a model reading the transcript cannot miss it, because it was never there.
It does not know what it cannot verify. Asked whether the boiler was replaced or whether the seller would take less, the correct behaviour is to say plainly that it does not know and put a call on your calendar. A voice that invents an answer to sound competent is worse than no answer at all, and worse on the phone than anywhere else, because the caller has no way to check it and every reason to believe a confident voice.
It does not pretend to be a person. Ours introduces itself as an assistant. In practice nobody minds. What people mind is nobody picking up.
And it does not replace ringing back the people who matter. If the caller is a past client, or your seller's neighbor, or anybody whose relationship is the actual asset, the agent's job is to hold the door open for twenty minutes. It is not to be the relationship.
The legal part nobody sells you
Nobody selling phone AI opens with this, which is a decent reason to read it here. None of it is legal advice and all of it moves. The point is to know that the questions exist, because the person selling you the agent is not going to raise them.
An AI voice is not a loophole. In a declaratory ruling adopted on February 2, 2024 and released six days later, the Federal Communications Commission confirmed that calls made with AI-generated voices are "artificial" for the purposes of the Telephone Consumer Protection Act. In practice that means the rules already governing prerecorded outbound calls, getting consent, identifying who is calling, and honoring an opt-out, apply in exactly the same way when the voice is generated. Dialing people with an AI sits inside the existing rules rather than around them.
Recording is a state question, and New York is not the strict one. The offense is Penal Law section 250.05, which makes a person guilty of eavesdropping "when he unlawfully engages in wiretapping, mechanical overhearing of a conversation, or intercepting or accessing of an electronic communication". The one of those three that governs a telephone call is defined in section 250.00: wiretapping means recording "a telephonic or telegraphic communication by a person other than a sender or receiver thereof, without the consent of either the sender or receiver". If you are on the call you are one of those two, so in New York you may record your own calls without asking the other side. California is the other kind of state: Penal Code section 632 makes it an offense to record a confidential communication without the consent of all parties, and several other states take the same approach. Before you lean on any of this, check the rule in the state your caller is actually sitting in, because the lists of these that circulate online go stale and none of them are the statute. If you list in the Hudson Valley and take a call from somebody in Los Angeles, assume the stricter rule.
Take this to anybody, including us
Three questions that separate a product from a demo.
Ask what it says in its first sentence.
If the answer is a person's name rather than the word assistant, stop there. Everything else on this list is downstream of whether the vendor treats disclosure as a feature or as a problem to be managed.
Ask what it does with a question it cannot answer.
You are listening for one specific behavior: it says so, and it books a call. If the demo never once shows the agent failing at something, you have been shown an advertisement rather than a product.
Ask to hear a call that went sideways.
Anything sounds good reading a script to a cooperative caller. Ask for a recording where somebody interrupted it, changed their mind halfway, or went quiet, because that is roughly every third real call.
Some states already require the disclosure you should be making anyway. California requires that a person called be told when a prerecorded message uses an artificial voice, defined as a voice generated or significantly altered by artificial intelligence. That is Assembly Bill 2905, chaptered in September 2024.
Our position is simpler than the map. The agent says that it is an assistant and that the call is recorded, at the start of the call rather than in a rushed sentence at the end. Not because a statute in another state says so, but because there is no version of this where being told costs you more than being found out.
And if you are running outbound at any volume, have your own attorney read the script once. It is an hour, and it is the cheapest hour in the whole project.
The script itself
Ask and we will send you in writing the exact opening our agents use: the line that says it is an assistant, and the line that says the call is recorded. Give that to your attorney rather than a blank page.
One page, no charge, and yours to use whoever ends up building the thing.
The system
What it is connected to.
An agent is only as good as what it can reach. This is every hop between the ring and the appointment you walk into.
Scroll to follow the chain
- The call: Your own number, any hour
- The agent: A voice on Vapi, given limits
- Your listings: What it may state
- Your calendar: Live availability, not a guess
- The record: CRM, with the transcript
- The handoff: Time booked with a person
Where it goes wrong
Three failure modes. None of them are the technology, which is exactly why they are the common ones.
Three ways it fails
None of them are the technology.
It is set up to answer everything.
An agent that never hands off eventually becomes the thing standing between a serious buyer and you. The most important setting is not what it can say. It is the point at which it stops talking and books.
Nobody told it what it may not say.
The limits are the configuration. Give an agent no boundary on price, condition or anything legal and it will fill the silence, fluently, and you will hear about it from the client rather than from the log.
The transcripts are never opened.
Every call is a record of the exact questions your market is asking, in their own words, with the objection attached. It is the best sales material you will ever own and most people never read a line of it.
Common questions, answered honestly
Will callers know they are talking to an AI?
Yes, because it tells them. It introduces itself as an assistant at the start of the call. In practice the objection is far smaller than people expect: callers care enormously that somebody picked up at nine on a Sunday and answered the question, and very little about who. What we will not build is an agent that pretends to be a particular human being.
What happens when it cannot answer something?
It says so plainly and books time with you instead of guessing. It will not invent a price, a legal position, or a fact about a property. Anything it cannot verify turns into a booked call, which is both the honest outcome and the higher-converting one.
Can it replace an inside sales agent?
It replaces the mechanical half of the role: answering every call, dialing new leads within seconds, qualifying on budget, area and timeline, and booking the appointment. It does not replace an experienced agent at a listing appointment. Most teams run the AI as first contact and keep people for the conversations where judgment and relationship decide the outcome.
Does it work with my calendar and my CRM?
Yes. It reads your live availability, so it only ever offers a time you actually have, and it writes the transcript, the qualification and the outcome back to your CRM. The appointment booking side is the same machinery. It plugs into what you already run rather than adding one more thing to check.
Is it legal to have an AI make outbound calls?
The same rules that govern your own outbound calling apply: you need consent to call, you have to honor do-not-call requests, and the agent identifies itself as an AI assistant. The FCC settled the underlying question in February 2024 by ruling that an AI-generated voice counts as an artificial voice under the TCPA, so nothing about using one relaxes the rules you were already under. We configure calling windows and opt-outs to match, and every call is logged so the record exists.
What to do about it
There is a test for this that takes ninety seconds and costs nothing, and it is worth doing before you spend a dollar with anybody, us included. Borrow a phone your office does not have saved, ring your own number on a Sunday evening, and sit through the whole thing the way a stranger would. Count the rings. Listen to what your voicemail actually says, all the way to the beep. Then decide whether that is the first thing you want somebody standing outside your listing to hear.
Most people who do that come away having learned something they did not want to know, which is the point. You can hear the alternative on the RealtyLT AI page. The AI voice agents page has the rest of it: what the agent is wired to, and what it does with a call once it has taken one.
The same argument at the other end of the day is on the website chat side, where the buyer messaging you at 11:40pm never picks up the phone at all.
The buyer who rang at 9:42 is not going to try you again on Monday. They are being shown a house on Saturday by whoever picked up. The only question left is whose listing they are standing in.
The chat assistant on that page is live and will answer you right now. A voice agent is stood up per office, against your own number and your own calendar, so there is no shared demo line to ring.
