Design the handoff before writing more bot replies
A WhatsApp chatbot for real estate can collect a useful first detail, but a customer still needs a clear path to a person when the question needs judgement, confirmation or account access. For a Dubai agency, the handoff should preserve the property reference and the customer’s actual question so the agent does not ask them to repeat the conversation. This guide offers a testable workflow to discuss with a developer. It does not claim that Devzonia currently offers a particular bot feature or integration.
Choose clear reasons to stop automation
Write a short list of handoff triggers before drafting canned replies. Examples include an unconfirmed price or viewing slot, a complaint, a request to change a reservation, a question about personal documents, or a customer explicitly asking for a person. Make the rules visible to staff. A bot should not keep rephrasing an answer when the approved information is missing. It should say what is unknown, offer a human route and retain the unresolved question.
Define the context an agent needs
For each handoff, pass only information relevant to the request: the WhatsApp conversation reference, language selected by the customer, property reference if supplied, the exact question, facts already checked, and the reason automation stopped. Do not add inferred budget, nationality, urgency or intent unless the customer supplied it and the agency has a lawful reason to use it. Make the next owner explicit, such as an assigned agent or a monitored queue, and identify who watches unassigned cases.
Map four simple routes
| Situation | Bot response | Agent context | Check |
|---|---|---|---|
| Missing service charge | State it is unconfirmed and offer a check | Property ID and missing field | No amount is invented |
| Viewing request | Collect a preferred time without promising a booking | Property ID, requested time and contact reference | Appointment remains pending until confirmed |
| Customer requests a person | Acknowledge and route promptly | Conversation reference and original request | No repeated qualification loop |
| Complaint or sensitive issue | Pause scripted sales questions and offer the agreed support path | Customer’s words and timestamp | Access limited to responsible staff |
Test the route as a customer and as an agent
Use fictional records and a test number. Run each route in Arabic and English, include a message outside business hours, and try a customer who changes their mind. Confirm that the customer sees a clear acknowledgement, the agent receives the right context, and the case stays visible until it is accepted or closed. Record the exact prompt, expected behaviour, observed reply, owner and retest. Do not use real customer data in a demo unless access and privacy controls have been approved.
Set expectations that the agency can meet
A handoff message should explain what happens next without promising a response time the team has not staffed. If the queue is unmonitored at night, say that plainly and provide the approved alternative. Decide how a case is returned to automation after a human reply and what happens when two agents respond. These are operating choices that need agreement; a chatbot cannot create staffing coverage or guarantee an agent will accept every case.
Brief a scoped implementation
Ask a provider to demonstrate the four routes with sample data and show where the transcript, owner and unresolved question are stored. Clarify whether WhatsApp connectivity, CRM updates, Arabic handling and human routing are included, require a third-party service, or would need custom development. Ask about permissions, retention and failure alerts. For a project discussion, take this map to Devzonia and request a written scope and acceptance criteria. Treat the published service information and demonstrated behaviour as the evidence; do not assume a portal partnership or existing feature without confirmation.
Use the article alongside the agency’s main real estate CRM requirements. Suggested internal destination: the core CRM service page, once its live URL and supported capabilities are confirmed.
