Skip to content

Chatbot for real estate: what to automate first

· Automation · · 4 min read

Most enquiries an estate or letting agency receives ask the same handful of things: is it still available, what does it cost, can I see it, and is it right for me. A chatbot can answer much of that at any hour. It can also do damage if it guesses. This note sets out the approach we would take, starting with what to automate first and what to leave to an agent.

Answer from the listing data, not from the model

Price, availability, bedrooms, parking, pets and service charges should come from the listing record the agency already maintains, not from whatever a language model infers from a photo or a description. When a field is empty, the chatbot should say it does not know and offer to ask the agent. A confident wrong answer about a lease length or a fee is worse than no answer.

That makes the listing data the real project. If a property is let or under offer and the record still says available, the chatbot will say available. Before launch, agree who updates status and how quickly, and treat that update as part of the workflow.

Qualify without interrogating

Agents want to know a buyer's budget, timing and whether they need to sell first; a tenant's move-in date and who will live there. A chatbot can ask for these naturally while answering questions, and pass them on with the enquiry. Keep it to the few details the agent actually uses to decide what to do next. A long questionnaire before a single answer drives people away.

Book viewings against the real diary

A viewing request is only useful if the slot exists. The chatbot should offer times from the agent's actual calendar, respect the rules the office sets, such as no viewings before a property is ready or a buffer between appointments, and send a confirmation. Where an office prefers to confirm by phone, the chatbot can collect preferred times and hand over instead of booking.

Hand these to an agent, always

  • Offers, negotiation and anything about price flexibility.
  • Questions about contracts, deposits, legal obligations or eligibility.
  • Complaints, maintenance problems and anything urgent from an existing tenant.
  • Any question the listing data does not answer.

Each hand-off should arrive with the conversation so far, the property, and the details already collected, so the agent does not start again.

Test before customers do

  • Ask about a property that was let yesterday.
  • Ask for a viewing at a time the office is closed.
  • Ask whether the price is negotiable.
  • Ask about a property the agency does not list.
  • Send the same request twice from the same person.

The useful measure is not how many conversations finish without a person. It is whether every answer can be traced to the listing record and every hand-off reaches the right agent with what they need.

Planning a chatbot for your listings? AI chatbot development