Real estate buyers ask the most questions in the evening and at weekends, exactly when the sales team has the fewest people on duty. A question like "How much is a 2-bedroom unit?" left unanswered for 10 minutes is often a lost lead. A real estate AI agent covers exactly that gap and hands sales a lead with all the details already collected.
What a real estate AI agent does for agencies and developers
1. Answers from project documents, never invents. Price lists, sales policies, legal status, construction progress, location and amenities are loaded into the assistant: as files, the project website or a Google Drive folder. The assistant only answers within that scope. If a price has not been published, it says so and promises that sales will follow up, rather than giving a guessed number.
2. Asks for contact details at the right moment. When a customer asks about prices, payment plans or a viewing, that is the moment to ask for a phone number. The assistant does not ask in the first message. The number the customer leaves is masked before it reaches the AI model and is shown in full only in the CRM, to people with permission.
3. Pushes leads into the CRM with context. When a customer leaves a number, the assistant creates a customer record in the real estate CRM, tagged with the project, needs (number of bedrooms, budget range) and source channel, and creates a call-back task for the sales owner. When sales call back, they already know what the customer wants.
4. Present on every channel. The same assistant answers on the project website, Facebook page, Zalo OA and WhatsApp. See multichannel connections.
5. One assistant per project, per brand. Each project website has an assistant with its own colours, name and greeting. For example, on h-holding.vn the assistant greets with "Which project are you interested in?" and suggests prices, legal status and viewing schedules.
A sample conversation
Customer (21:40): Are there any 2-bedroom units left, and how much are they?
Assistant: The project still has 2-bedroom units of 68–72 m², priced from … according to this phase's price list (quoted exactly from the loaded price list). Which orientation would you prefer?
Customer: South-east facing. I'd like to visit this weekend.
Assistant: Certainly. So that our sales team can book a weekend viewing for you, may I have your phone number?
Customer: 09xx xxx xxx
Assistant: I have noted the weekend viewing and passed it to the sales owner, who will call you back tomorrow morning.
Behind the scenes: a new customer in the CRM (project, 2 bedrooms, south-east facing, wants a weekend viewing) and a "Call back – book viewing" task on the sales rep's task board.
Measurable results
The Chat Overview screen shows, per channel: how many conversations, how many genuinely interested customers (2 or more questions), how many left a number, how many became orders, and the conversation-to-order rate. You know which channel brings good customers, not just many customers.
What the AI Agent does NOT do
- It does not promise discounts beyond the loaded policy.
- It does not make legal claims beyond the documents.
- It does not guess a customer's gender from their name; it uses a neutral form of address until the customer says how they wish to be addressed.
- It does not replace sales in closing the deal. It keeps hot leads warm and passes them to the right person.
Frequently asked questions
What if the price list changes constantly?
Update the file or the Drive folder. The assistant reads the new version after it is reloaded, and the admin screen clearly shows which sources have finished loading.
Can several projects share one assistant?
Yes, or each project can have its own assistant if each has its own website and brand.
Will leads be duplicated across channels?
Customers are identified by phone number and email. The same person messaging through several channels ends up in a single profile.
Running ads for a project but losing leads after hours? Contact Le Anh – 0388.610.885 to set up a trial assistant for one project using your own sales documents. See also Agentic AI for business.
