Key takeaways
- A press-1 menu makes callers sort themselves before they say a word; conversational answering asks what they want and sorts it for them.
- Menu exits are expensive: 80% of Americans say they don't generally answer calls from unknown numbers, so a caller who hangs up is hard to win back[1].
- Menus are quick to build and slow to keep accurate. Conversational answering needs written routing rules, a current listing sheet and someone reading transcripts.
- Keep digits for choices callers can make reliably, such as language or a known extension, and let everyone else talk.
The menu was built around the org chart
Most brokerage phone menus were written on the day the phone system went in, by whoever was setting it up, and they describe the office the way the office sees itself. Press 1 for listings. Press 2 for the agent directory. Press 3 for property management. Press 0 for the front desk. Callers do not arrive with a department in mind. They arrive with a question: what is the house on Linden Street listed for, can I see it Saturday, my neighbor sold last month and what would mine get.
That mismatch is the whole comparison. A press-1 menu asks the caller to sort themselves before they have said a word. Conversational answering, whether by a person at the desk or an SI Employee, lets them say what they want and does the sorting for them. Almost everything else follows from which side sorts the call: how many callers give up, how often a call lands on the wrong agent, and how much work it takes to keep the setup right.
This guide compares the two on those three points for a brokerage main line. The plumbing underneath, ring groups, extensions and forwarding, is covered in brokerage phone systems: lines, routing and extensions.
Where callers drop off
Menus lose callers in three places. The first is the greeting: a welcome, a recording notice, a line about the website, and only then the options. The second is the choice itself. A buyer calling off a yard sign does not know whether a listing question belongs under option 1 or with the listing agent under option 2, and they rarely know the agent's name anyway. The third is the end of the branch, which at most brokerages is a voicemail box.
Each of those exits is expensive, because a caller who hangs up is hard to get back. In Pew's 2020 survey, 80% of Americans said they don't generally answer their cellphone when an unknown number calls, and only 19% said they generally do[1]. The agent's callback comes from a number the caller has never seen. Sellers in particular rarely shop around: 67% of agent-using sellers contacted only one agent[2]. The one they reach first has the edge.
- Generally answer 19%
- Don't answer, check voicemail 67%
- Don't answer, ignore voicemail 14%
Source: Pew Research Center, 2020[1]
Show the data
| Item | Value |
|---|---|
| Generally answer | 19% |
| Don't answer, check voicemail | 67% |
| Don't answer, ignore voicemail | 14% |
Automated systems also fail in a way callers have learned to expect. Looking at chatbots in banking, the Consumer Financial Protection Bureau described “doom loops”: a customer whose issue falls outside the system's limited options goes round in circles with no offramp to a human[3]. A phone menu without a zero-out option is the same trap with touch tones. Callers respond the way residents did in the property management example: they learn which digit reaches a person and press it for everything, and the menu stops sorting at all.
Conversational answering has exits too. Some callers hang up on hearing that an automated assistant answered, and some ask for a person straight away. Keep those exits small with a short greeting that says plainly what the caller has reached, an open first question such as “Which property are you calling about?”, and a real path to a person when someone asks for one.
Routing accuracy: digits versus answers
A menu routes on one keypress, so it is only as accurate as the caller's guess, and the guess is worst on the calls a brokerage most wants. A sign caller wants the listing agent but knows an address, not a name. A seller wants “someone who does valuations”, which is not an option. A cooperating agent wants showing instructions that differ by listing. Each of them presses something, and a share land on the wrong extension, ring out, or roll to a mailbox nobody checks on Saturdays.
Conversational answering routes on what the caller says. An SI Employee set up for a brokerage asks which property or area the call is about and routes by listing or team rule: the listing agent for a listing question, the area list or the round robin for a new seller, the office for everything else. When the right person is free it transfers the call; when nobody is, it takes a detailed message for them. Listing questions are answered only from the listing sheet the office approves, and anything not on the sheet is marked “not confirmed” for the agent. A forwarded call does not reveal which sign or ad the caller saw, so it asks.
That accuracy has a cost of its own: the routing rules have to exist in writing. Plenty of offices discover during setup that their real rule for new leads is “whoever is standing near the desk”. Writing the rule down helps whichever system you choose, and lead routing rules for real estate teams walks through the usual options.
| What you compare | Press-1 phone menu | Conversational answering |
|---|---|---|
| Who sorts the call | The caller, by guessing a digit | The line, from what the caller says |
| Typical exits | Hang-up in the menu, wrong extension, voicemail | Hang-up at the greeting; message when nobody is free |
| Listing questions | Not answered; sent to an extension | Answered from the approved listing sheet or marked not confirmed |
| New seller calls | Wherever the seller guesses | Address, timeline and plans taken; consultation offered |
| Evenings and weekends | Same menu, mostly ending in voicemail | Same questions as daytime; message or booking |
| Setup | Record prompts, map digits to extensions | Approve greeting, listing sheet and routing rules; run test calls |
| Upkeep | Re-record when options or agents change | Update the listing sheet and rules; read transcripts |
| Record of the call | Call log entry; a voicemail if one was left | Recording, transcript and summary for every call |
Setup and upkeep, honestly compared
A menu is quick to build. Most hosted phone systems let you record three or four prompts, map each digit to an extension or ring group and set an after-hours schedule in an afternoon. Upkeep is where menus decay. Agents join and leave; the median REALTOR has been with their current firm for six years[4], so even a stable 30-agent office changes its directory several times a year. Each change means editing a prompt, and in practice the menu ends up describing the office as it was two years ago.
Conversational answering takes longer to set up, because you have to write down what the office actually knows. That usually comes to four documents. The greeting, with an automated-assistant disclosure and a recording notice where your state requires one. A listing sheet with price, status, beds and baths, HOA basics and showing instructions for each active listing. The routing rules. And the boundaries: no pricing opinions, no descriptions of who lives in a neighborhood, and never reading out or confirming wire instructions. Then test calls before any real caller arrives, at minimum a live listing, a withdrawn listing and a showing request.
The upkeep is different in kind. Nobody re-records anything, but somebody owns the listing sheet and changes it the day a price drops or a listing goes pending, and somebody reads a sample of transcripts every week. If those two jobs have no owner, conversational answering decays as well, only more quietly than a menu. Testing an AI phone assistant before launch has a test-call script to start from.
What callers think of automation
Callers know when software is answering, and many are wary of it. In a Pew survey fielded in December 2022, 65% of US adults correctly identified customer service chatbots as a use of AI[5]. By 2025, 50% of US adults said they were more concerned than excited about the growing use of AI in daily life, up from 37% in 2021[6].
Source: Pew Research Center, 2025[6]
Show the data
| Item | Value |
|---|---|
| 2021 | 37% |
| 2025 | 50% |
The same Pew research shows what reassures people: 76% say it is extremely or very important to be able to tell whether content was made by AI or by a person[6]. On the phone that becomes three rules. Say at the start that an automated assistant is answering. Make the way to a person obvious, and honor it. Never let the line guess, which is why unconfirmed listing details go back to the agent. Disclosing AI to callers and leads has greeting wording.
Menus get no pass on this. A caller pressing 0 in a menu that loops back to the start is in the same trap the CFPB described[3], without even the chance to explain what they need.
When a menu is still the right tool
A menu is a good answer to a narrow question. When callers can sort themselves reliably, one digit is faster than a conversation.
- Language choice. “Para español, oprima dos” works because every caller knows which language they speak. Send that branch to a bilingual agent or ring group.
- Known extensions. Clients in the middle of a transaction often know their agent's extension. Let them dial it at any point during the greeting.
- Staff-only lines. Agents calling for the broker on duty can have their own number or digit, as described in a broker on-call line for agent questions.
- Very low volume. An office that takes a handful of calls a day, nearly all for agents by name, may be better served by a dial-by-name directory and a well-run ring group.
Menus fail at the sorting callers cannot do for themselves: new buyers, new sellers and anyone who does not know a name. That is most of the new business on a brokerage main line, and it is where conversational answering earns its setup time.
A hybrid many brokerages land on
The usual end point is not either-or. Keep one fast path for callers who know where they are going and let everyone else talk. In practice the greeting offers extension dialing for a couple of seconds and then hands the call to conversational answering if nothing is pressed.
- Caller
- System
- SI Employee
- Your team
01Caller
Buyer calls the office
Saturday 11:15 am, parked outside a house with the brokerage sign.
02System
Short greeting
Dial an extension now or stay on the line, with disclosure and recording notice.
03SI Employee
Asks which property
Answers price, status and showing basics from the approved listing sheet.
04SI Employee
Takes a showing request
Preferred times, and whether the buyer already works with an agent.
05SI Employee
Routes by listing rule
Transfers to the listing agent if free; otherwise takes a detailed message.
06Your team
Agent confirms the time
Summary, recording and transcript arrive by email and webhook.
Example flow. The showing stays a request until the listing agent confirms it.
Two settings decide whether the hybrid works. The first is how long the greeting waits for a digit: two or three seconds is plenty, and much longer starts to feel like a menu again. The second is what happens when a transfer is not answered. The caller should come back to the line, hear plainly that the agent is not available and leave a message that is flagged urgent if it meets your rule, instead of ringing out into a mailbox. How to transfer a call without losing the caller covers the transfer side in detail.
What menu drop-off can cost
Most hosted phone systems can report calls that ended inside the menu, broken down by option. That report is the input the worked example below needs. If yours cannot, a week of call logs set against voicemails and logged leads gives a rough figure.
| Input | Example |
|---|---|
| Calls a week to the main line | 150 |
| Share that hang up in the menu or at voicemail | 8% |
| Share from new buyers or sellers | 20% |
| Share of those that become a closed side | 3% |
| Gross commission per closed side | $9,000 |
| Unanswered calls a month | 52 |
| From new customers | 10 |
| Bookings that would have followed | 0.3 |
| Monthly value at risk | $2,806 |
How it is worked out: Calls a week × 4.33 weeks × unanswered share × new-customer share × booking rate × average value.
Assumptions
- Value per side: $9,000 gross commission, about 2.2% of the $417,500 median existing-home price in September 2026. Use your own figure.[7]
- Lost callers stay lost: 80% of Americans say they don't generally answer calls from unknown numbers, so a callback rarely rescues a hang-up.[1]
- The 8% drop-off, the new-business share and the close rate are example figures. Take yours from the phone system's menu report and your CRM.
Example only. Monthly commission at risk = calls a week × 4.33 × drop-off share × new-business share × close rate × value per side.
In this example about $2,800 a month in gross commission rides on callers who give up in the menu, roughly one closed side every three months. Check it against your own numbers before relying on it, and notice which way the error usually runs: the callers most likely to quit a menu are the ones who don't know whom to ask for, which means new prospects rather than existing clients.
Test both on your own line for a week
Pull the menu report
Count calls that ended in the menu or in a voicemail box over the last 30 days, by option. Note which option gets pressed most; it is usually the one that reaches a person.
Call your own line as five callers
A sign caller who knows only the address, a seller, a cooperating agent, a tenant in a listed rental and a vendor. Time how long each takes to reach the right person or leave a useful message.
Write the routing rules down
Named agent, listing agent, area list, round robin, office. If they don't fit in ten lines, neither a menu nor a conversation will route well.
Run conversational answering on overflow first
Forward only unanswered and after-hours calls for a week, keep the menu for everything else, and compare outcomes call by call.
Judge on outcomes
Count calls that reached the right agent or ended in a complete message, not calls answered. A line that answers everything but routes badly has only moved the problem.
If you go ahead, the live list of what an SI Employee handles on a call today, and what is still planned, is on what it does. Read handing a call from the line to a human agent before you set the transfer rules.
Frequently asked questions
Is an IVR the same thing as an auto attendant?
In everyday use, close enough. An auto attendant is the greeting-and-menu feature of a business phone system; IVR is the wider term for systems that respond to keypresses or simple spoken commands. Both ask the caller to sort themselves, which is the part that matters in this comparison.
Can we keep the menu and add conversational answering only after hours?
Yes. Forward the line on a schedule for evenings and weekends, or only when calls go unanswered. Many offices start there because the old voicemail box gives them something easy to compare against.
Do callers hang up when they hear an automated assistant?
Some do, and wariness is real: 50% of US adults were more concerned than excited about AI in daily life in 2025[6]. A short greeting that says what the caller has reached, an open first question and an easy path to a person keep that exit small.
What should happen when a caller asks for a person?
Put it in the routing rules. In staffed hours, transfer to the front desk or the duty agent. When nobody picks up, the caller should hear that plainly and leave a message with a named owner, never hit a dead end.
Does conversational answering know which listing the caller saw?
Not from a forwarded call. It has to ask which property or address the call is about, then answer from the listing sheet you approve. Keep the sheet current or the answers go stale.
References
Pew Research Center (2020). Most Americans don't answer cellphone calls from unknown numbers (opens in a new tab) ↑
Zillow Research (2026). Sellers: Results from the Zillow Consumer Housing Trends Report 2026 (opens in a new tab) ↑
Consumer Financial Protection Bureau (2023). Chatbots in consumer finance (opens in a new tab) ↑
National Association of REALTORS® (2026). Experienced REALTORS® Anchor the Industry as Housing Affordability Remains Top Hurdle (2026 Member Profile) (opens in a new tab) ↑
Pew Research Center (2023). Public Awareness of Artificial Intelligence in Everyday Activities (opens in a new tab) ↑
Pew Research Center (2025). How Americans View AI and Its Impact on Human Abilities, Society (opens in a new tab) ↑
National Association of REALTORS® (2026). Existing-Home Sales Increased 1.5% in September, According to NAR Report (opens in a new tab) ↑