Real estate enquiries die in the gap, not in the team. A study reported in Harvard Business Review found that firms that tried to contact a lead within an hour were nearly 7 times as likely to qualify it as those that waited even an hour longer. The team is not the problem. The gap is.
This post is the script template plus the inbound voice agent setup we use to close that gap in 90 seconds - what we call The 90-Second Inbound Loop.
- 90 seconds from form-submit to a real-sounding voice on the lead's phone.
- Script structured as 6 fixed beats: confirm, compliment, qualify x2, book, recap.
- Stack we deploy: Vapi + ElevenLabs + Make.com + Google Calendar + Twilio.
- Want to know where your inbound leaks before you spend anything? Start with the free 30-min audit call.
1. Why the script matters more than the model
Most operators evaluating voice agents start by comparing models. ElevenLabs vs PlayHT. GPT-4o vs Claude. They run the demos, hear something stilted, and decide voice AI is not ready. The demo is not the bottleneck. The script is.
A great model with a bad script sounds like a chatbot reading a brochure. A mid-tier model with a tight script sounds like your best SDR on day 30. The difference in booked calls comes from a few hundred lines of script logic, not a different vendor.
This is why every inbound voice agent we build as part of a sales system starts with the script. The model gets swapped if and when it improves. The script earns the bookings.
2. The 90-Second Inbound Loop - six beats
The loop has six beats and exactly six. Skip a beat and the lead disengages. Add a beat and you slow the call past the 90-second sweet spot.
| # | Beat | Time | Job |
|---|---|---|---|
| 1 | Confirm | 0:00 - 0:08 | Confirm the right person and the property they enquired about. Drop hesitation. |
| 2 | Compliment | 0:08 - 0:18 | Acknowledge the choice. One specific detail about the property. Builds trust fast. |
| 3 | Qualify A | 0:18 - 0:35 | One question on timeline ("Are you looking to view this month or are you exploring?"). |
| 4 | Qualify B | 0:35 - 0:55 | One question on financing or use ("Self-funded, mortgage, or investor? Live in or rent?"). |
| 5 | Book | 0:55 - 1:20 | Offer two specific slots. Default to "today or tomorrow", never "this week". |
| 6 | Recap | 1:20 - 1:30 | Repeat day, time, and how the agent will arrive. Confirm SMS sent. |
Total: 90 seconds. Less is too cold. More and you lose the ones who answered while distracted.
3. The actual script template
Below is an example script for an off-plan developer selling to international buyers, with the qualify questions tuned to them. Swap the project name and the specific compliment line for your inventory and you have a working v1.
[BEAT 1 - CONFIRM]
"Hi, is this {firstName}? This is Sara from {projectName}. You sent us an enquiry about the {unitType} on {projectName} {minutesAgo} minutes ago. Do you have 90 seconds?"
→ If "no": "No problem. When today works - in an hour, or this evening?"
→ If "yes": continue.
[BEAT 2 - COMPLIMENT]
"Quick note before I take your details - that {unitType} is one of three with the {specificFeature}. Most enquiries are coming in for the standard layout, so good eye."
[BEAT 3 - QUALIFY A: TIMELINE]
"Are you looking to view this month, or earlier-stage exploring?"
→ "this month": flag as HOT, prioritise booking today/tomorrow.
→ "exploring": continue but offer remote video walkthrough as primary.
[BEAT 4 - QUALIFY B: FINANCING / USE]
"And quick one - self-funded, mortgage, or investor? And live-in or rental?"
→ Capture both. Routes to right agent in CRM.
[BEAT 5 - BOOK]
"Perfect. I have two slots that fit:
- Tomorrow at 11am, on-site walkthrough with {agentName}.
- Today at 4pm, video walkthrough with {agentName} from your phone.
Which works?"
→ If neither: "What time today or tomorrow does work?" - never offer "this week".
[BEAT 6 - RECAP]
"Booked. {agentName} will {arrivalMethod} at {time} on {day}. You'll get an SMS in the next 60 seconds with the address and a calendar invite. Anything else before I let you go?"
Three rules on the script:
- Never let the agent improvise outside these beats. Improvisation is what makes voice agents sound robotic. Tight rails sound human.
- Default to today or tomorrow. "This week" is the kiss of death. Specific beats vague.
- End on a question. "Anything else?" closes the loop without leaving the lead hanging.
4. The seven mistakes that kill voice agent ROI
The same mistakes show up in most real estate voice agent setups. The five biggest get a full deep-dive in 5 voice-agent failure patterns that kill real estate ROI; this post covers two more.
- Three-question qualifying. Two is the cap. Three is an interrogation.
- "How can I help you today?" Never. The agent already knows. State the property and move.
- No backup slot. One slot equals a dead booking. Always two, then "what works for you" as fallback.
- Voice tuned for charisma instead of clarity. Calm and clear beats warm and bouncy on a phone speaker.
- No SMS confirmation in the same flow. If the lead does not get an SMS within 60 seconds of the call ending, no-show rate doubles.
- Calling at 11pm because Make.com runs 24/7. Set call windows. 9am to 8pm in the lead's timezone.
- No human escalation path. The agent must hand off to a real number for any "I want to speak to someone" or budget-clarifying question.
5. Common misconceptions
Three beliefs cost real estate teams the most when they evaluate voice agents. Address them before you scope a project.
- "Voice agents sound robotic - my buyers will hang up."
- The default voices in vendor demos are not the production state. With a well-chosen voice and a few weeks of script tuning on real calls, most callers stay in the conversation. The ones who notice it is AI mostly want to be sure they reached a real company, so the agent should say so plainly when asked.
- "AI replaces our SDR team."
- It does not. The agent qualifies and books in 90 seconds. The human SDR picks up a calendar-confirmed lead with two qualifier answers already captured. Voice agents extend coverage to 24/7 and remove the 17-minute first-call gap. They do not handle budget objections, contract questions, or "I want to speak to a person" intent - those still need a human.
- "You need a custom-built solution to handle real estate properly."
- You need three things: a tight 6-beat script, a voice that matches your brand, and a clean handoff to your CRM and calendar. The same Vapi + ElevenLabs + Make.com stack works across very different businesses, a pattern documented further in McKinsey's State of AI, which finds vertical specialisation outperforms custom-builds for mid-market AI deployments. Custom-built voice infrastructure is a 6-month project that mid-market operators do not need.
6. The math - voice agent vs human SDR
The case for voice over a human SDR is not "AI is cheaper". It is "AI is faster, and speed is what closes real estate enquiries". Here is the comparison we run with operators on the call.
| Human SDR (in-house) | Inbound voice agent | |
|---|---|---|
| Time to first call | Office hours, when someone is free | 90 sec |
| Coverage | 9am - 6pm, Mon - Fri | 24/7 within set windows |
| Capacity | ~40 calls/day | Unlimited concurrent |
| Running cost | Salary, hiring time and ramp | Platform minutes plus telephony; the build is quoted after the audit |
| Books in real-time | Manual to calendar | Direct write to Google Calendar |
The voice agent does not replace the SDR. It buys you time before the SDR is needed. A well-built agent hands the qualified, calendar-confirmed lead to a human. The human picks up a warm conversation, not a cold dial.
7. How the build runs
The inbound voice agent is one part of a sales system, and the scope comes out of the free audit. The work itself follows the same order every time:
- Script workshop. We map your inventory, your pricing tiers and your routing rules. Output: a signed-off script in the 6-beat format.
- Voice tuning. Pick the voice by testing a short list against your brand. Lock pacing, fillers and recovery patterns.
- Stack wiring. Form-submit routes to the voice platform, telephony and your calendar. The SMS confirmation flow goes live.
- QA. Internal test calls and your own test calls. Anything that fails the 6-beat checklist gets rewritten before launch.
- Supervised first run. The agent goes live on real inbound leads, and the founder hears every call recording while it settles.
The go-live date is agreed in writing once access and review needs are known. If a vendor cannot commit to a date in writing, the script is not tight enough. That is the diagnostic.
What to ship this week
If you are losing leads in the gap between the form and the first call, the next move is not "compare three voice vendors". It is to find out what your loop is actually losing in dollars per week. Book the free 30-min audit call and we will look at your live form, your CRM and your recent enquiries together, then tell you where the inbound leaks.
The agencies that win the next 12 months in real estate are not the ones with the prettiest sites. They are the ones whose voice agents pick up at 11pm on a Sunday when a buyer in Singapore decides to enquire about a Canggu villa. That call gets answered in 90 seconds, or it does not get answered at all.
Frequently asked questions
What is the ideal length of a voice agent script for inbound real estate enquiries?
90 seconds, structured as 6 fixed beats: confirm the lead and property, compliment one specific feature, ask one timeline question, ask one financing or use question, offer two specific slots (today or tomorrow, never "this week"), then recap with day, time, and SMS confirmation.
Vapi vs Retell vs ElevenLabs - which voice agent stack should real estate use?
Stack choice matters less than script quality. kratt defaults to Vapi as the orchestration layer with ElevenLabs voices for English-speaking international buyers and Twilio for telephony. Retell works equally well. Most of the difference in booked calls comes from the script logic, not the vendor.
How fast does a voice agent need to call back an inbound real estate lead?
Within 90 seconds of form-submit if you can, and inside the first hour at the latest. Harvard Business Review found that firms that tried to contact a lead within an hour were nearly 7 times as likely to qualify it as those that waited even an hour longer. Fix the response time before you tune the script.
Does a voice agent replace a human SDR in a real estate sales team?
No. The voice agent qualifies and books the calendar in 90 seconds. The human SDR picks up a warm, calendar-confirmed conversation. Voice agents extend coverage to 24/7 within set call windows and remove the 17-minute first-call gap. Human handoff is required for budget clarification and any "I want to speak to someone" intent.
How long does it take to deploy a voice agent for a real estate developer?
The scope comes out of the free audit, and the go-live date is agreed in writing once access and review needs are known. The build covers a script workshop and 6-beat sign-off, voice tuning across candidate voices, stack wiring (Make.com, Vapi, Twilio, Google Calendar, SMS), internal and client QA calls, and a supervised first run where the founder hears every call recording.
What is the cost of a voice agent vs hiring an in-house SDR for real estate?
Running costs are platform minutes plus telephony, and kratt prices the build after the free audit, once volume and integrations are known, so there is no rate card to quote here. An in-house SDR adds salary, hiring time and ramp. The agent does not replace the SDR. It buys time before one is needed and lets the SDR focus on warm conversations only.
Related resources
- Sales systems - one sales system from the first lead to the signed deal, with inbound call handling as one part.
- Case studies - client work with the receipts we can show.
- Business systems and integrations - how the voice agent writes into your CRM, calendar and reporting.
- Book the free 30-min audit call - we look at your live form and your recent enquiries together.
Last updated: 1 May 2026 · This guide is updated quarterly with new benchmarks and case data.