Build a solo agent's AI phone assistant that books buyer tours, logs showing requests and hangs up on spam
William Zhang, a real estate agent in Austin, Texas, writing on his own blog (Real Estate AI Society)
When he doesn't pick up, a voice agent named Ashley answers within two rings. A first step sorts the caller into buyer/seller lead, general inquiry (like another agent's showing request) or spam. Leads go to a scheduling path that checks his calendar, offers two time slots at a time and books the appointment with name and email. Showing requests and other inquiries become a detailed message with up to two follow-up questions (it doesn't give out lockbox or gate codes; it relays the request to him). Spam calls are ended immediately with no notification.
From $6/mo (checked prices)
Advanced · or we do it from $149/mo
ElevenLabs, Twilio, TidyCal
The problem
Calls came in during showings and video shoots. By the time he called back hours later, buyers had talked to someone else, and spam calls trained him to stop answering at all. Showing requests from other agents took 3–5 minutes each, five to ten times a week.
What happened
Self-reported: he checks the call log once a day and shows up to booked appointments; he estimates the showing-request calls alone used to take close to an hour a week. The examples in his post are test calls; no lead or deal counts shared. He puts typical voice-agent costs at $50–$200 a month depending on volume.
Quoted in the source: William Zhang, real estate agent, Austin, TX. Source: Real Estate AI Society (William Zhang): I built an AI voice agent to never miss a real estate lead again (2026-03-16) · Added 2026-10-14. Results as reported by the source; we haven't independently verified them.
3 new AI use cases every morning. Free. No spam, unsubscribe anytime.
How to set it up yourself
- What he used: ElevenLabs for the voice agent, a Twilio phone number, and TidyCal for calendar booking through its API.
- Forward unanswered calls (after two rings) to the agent's number so you still pick up when you can.
- Write a first 'sorting' instruction with three paths: buyer/seller lead, general inquiry, spam.
- For leads, connect your booking calendar and have it offer two open slots at a time, then collect name and email for the confirmation.
- For general inquiries, take a message and ask up to two follow-up questions; never read out access codes.
- Add edge cases as they happen (e.g. a neighbor asking about a listing goes to general inquiry).
- Review the log daily and call back showing requests that are time-sensitive.
Copy-paste prompt
You are Ashley, the AI assistant for [Agent name], a real estate agent in [City]. First decide which kind of caller this is: (A) someone who wants to buy, sell or tour a home, (B) a general inquiry such as another agent requesting a showing, or (C) a sales or spam call. For A: offer two open times from the calendar, book the one they pick, and collect their name and email. For B: take a detailed message, ask at most two follow-up questions that [Agent name] would need, and say they'll get a call back. Never share lockbox, gate or door codes. For C: politely end the call. Keep every reply short and friendly.
Tools used
ElevenLabs
AI voice generation for voiceovers and audio content.
From $6/mo (Starter) (checked 2026-10-13)
Visit site →Some tool links may be affiliate links. Disclosure
Checked prices: ElevenLabs: from $6/mo (Starter) (verified 2026-10-13)
Rather not do it yourself?
We'll set this up in your tools, in your voice, tested. $149/mo with setup included (6-month minimum), or $497 pay-once. Satisfaction guarantee: we'll fix it until it works, or refund your setup fee. Terms. We reply within 1 business day.
Get this set upRelated for Real estate
Have Claude sweep your CRM each morning and draft follow-ups for leads you'd forget
A Claude-based assistant pulls Follow Up Boss leads, notes and deals, scores each lead A–D using the Keller Williams model, drafts a pers…
Draft a month of real estate social posts with ChatGPT and repurpose the winners
Each evening ChatGPT drafted 20 to 30 short Threads posts, which he scheduled; the best performers became Canva quote cards, then 30 to 6…
Have AI draft replies to tenant maintenance requests and suggest the vendor, with staff approving
n8n workflows draft each reply to a tenant maintenance request, pull out details (unit, issue, photos, access), suggest a vendor type and…
Running something like this in your business? Submit your use case and we'll credit you with a link.
Get one proven AI setup a week
3 new AI use cases every morning. Free. No spam, unsubscribe anytime. Privacy