Two languages, seven requirements on every call: a bilingual AI voice receptionist for a Romanian real estate agency
A Romanian real estate agency needed every caller's property requirements taken down properly, in Romanian or English. Working white-label for an AI agency, we built an inbound AI voice receptionist that answers in Romanian, switches language mid-call, takes a seven-point property brief, reads it back and sends it to an agent for follow-up.
- Caller dials
- Voice agent
- Make.com
- Lead record
Snapshot
- Client type
- Real estate agency, delivered white-label through an AI agency
- Country
- Romania
- Industry
- Real estate
- System
- AI Voice Agent (inbound)
- Use cases
- AI receptionist · lead qualification
- Stack
- Voice AI platform · Make.com · Google Sheets or Airtable
- Timeline
- Delivered in April 2025
- Headline numbers
- 2 languages, switching mid-call · 7 property requirements on every call · contact details taken last
What did the client need?
Answer every call, take a complete property brief, and handle Romanian and English speakers naturally.What did we build?
A receptionist that asks one question at a time, takes contact details last and confirms every detail clearly in both languages.How does it run day to day?
Every call ends with a webhook; a Make.com automation writes the brief, summary, sentiment and transcript to the lead record.
What did the client need?
The brief came through an AI agency that sells AI systems to its own clients and outsources the build to us. Its client, a Romanian real estate agency, wanted every call answered, each caller's property requirements taken down, any relevant questions the caller missed asked, and the complete brief sent to a real estate agent for follow-up. It had to work naturally for Romanian- and English-speaking callers.
What did we build?
On every call the agent greets the caller, asks their name, then works through seven requirements one question at a time:
- Maximum budget
- Number of rooms
- Minimum surface area
- Preferred floor
- Parking space, yes or no
- Preferred neighbourhoods
- Fully equipped or just finished
It collects contact details last (full name, phone, email and a preferred callback time), reads the brief back for confirmation, tells the caller an agent will be in touch with matching listings, and ends the call.
Bilingual by design. The agent opens in Romanian and switches to English, or another supported language, as soon as the caller uses it. Each detail is confirmed in its own short sentence, phone numbers are read back digit by digit, and unclear names are spelled back letter by letter so the caller can correct them.
How does it run day to day?
Caller dials the agency ─► voice agent (Romanian / English) ─► end call ─► webhook ─► Make.com ─► lead record ─► agent follows up
Per call, the lead record holds the caller's details, the full property brief, call time and duration, call type, who ended the call, a summary, sentiment, a success flag and the full transcript. The agency's agents follow up from there with matching listings. If a caller strays off topic, the agent steers them gently back to the purpose of the call.
What were the results?
We don't have call volumes or conversion figures to publish. What the receptionist does on every call is fixed by design: it handles two languages and switches mid-call, captures the same seven property requirements in the same order, takes contact details last and confirms everything back before the call ends.
- Languages, switching mid-callBuild spec
- Property requirements captured on every callBuild spec
- LastContact details, taken after the property briefBuild spec
What did we learn, and what are the limits?
- One question at a time, contact details last. It keeps callers focused on what they want and produces a complete brief.
- Write explicit rules for numbers and names. Reading digits one by one and spelling unclear names makes confirmations unambiguous in both languages.
- White-label needs a complete brief. Requirements come through the agency, so every detail of the end client's process has to be in the brief, and the handover documentation has to be clear enough for the agency to support its own client.