A Bilingual AI Receptionist Pilot for Dubai Property Managers

An AI receptionist for property management in Dubai should prove that it can deliver a useful enquiry to the right person. A fluent greeting is only the opening step. The harder work is identifying the building, preserving the caller's language preference and separating a tenant's maintenance request from a prospective renter's question.

Dubai's rental activity makes this a practical workflow to examine: the Dubai Land Department reported 1.38 million registered tenancy contracts in 2025. That figure describes market activity, not the number of businesses that need an assistant. A pilot should still begin with one manager's actual call patterns. DLD rental report

Choose one portfolio and two call types

Start with a portfolio whose building names, staff responsibilities and operating hours are known. Select two tasks: capturing a leasing enquiry and recording a routine maintenance request. Keep live property availability, contract decisions and contractor dispatch outside the pilot unless the necessary systems are connected and tested.

For leasing, collect the preferred area, property type, budget, intended move date and callback details. For maintenance, collect the building, unit, reported problem and the caller's preferred contact method. Ask only for information the receiving team actually uses. An assistant that collects twenty fields but misses the unit number creates more work than it removes.

Record the proposed UAE phone route separately from the conversation design. Burki's UAE business page is a starting point for discussing the intended configuration; a regional page alone does not establish that a particular carrier arrangement is available.

Use local names in the test conversations

Prepare a building list with the spellings your staff recognizes. Include common spoken variants and the identifiers that distinguish similarly named properties. The assistant should ask a clarifying question when two buildings could match, rather than silently choosing one.

Consider this hypothetical call: an Arabic-speaking tenant gives a building name in English, then corrects the unit from 1406 to 1408. A successful result preserves the correction, records the final unit and leaves the problem description intact. Translating the conversation beautifully while retaining the wrong unit is a failure.

Have Arabic-speaking reviewers assess the spoken exchange and the resulting staff note. Ask an English-speaking reviewer to use that note without hearing the call. This exposes whether language conversion has lost practical details needed for the handoff.

Define what a finished enquiry looks like

Agree on a short acceptance record before testing:

Do not substitute “viewing booked” for “viewing requested.” Likewise, a recorded maintenance issue is not evidence that a contractor has been assigned. The assistant's closing sentence should match the action actually completed.

If HubSpot is the agreed destination, inspect Burki's HubSpot integration against the proposed contact-and-note workflow. Confirm which actions are enabled for that assistant instead of assuming access to every CRM object.

Review exceptions before expanding

Include a caller who switches languages, an unknown building, a corrected phone number, a repeated request and a caller who cannot provide a unit. Decide what staff should receive in each case. A partial but clearly labelled enquiry can be useful; a falsely complete record can send someone to the wrong property.

For urgent or unusual reports, use instructions approved by the property manager and a defined human escalation path. Do not let the assistant invent technical diagnoses, response times or assurances about a building's safety.

Judge the pilot by completed usable records, avoidable clarification calls and accurate ownership. Review failures individually before changing the script. Once both language paths work for one portfolio, extend the building list or call types one change at a time. Bring those two initial workflows to Burki so the pilot can be assessed against a concrete operational task.