Use AI to organise an authorised enquiry into an explicit brief, then apply clear rules for ownership and human review. Preserve the source, distinguish stated preferences from missing information and keep property availability in its current system.
Define qualified before assigning a score
Different agencies mean different things by a qualified lead. One needs a stated budget and a viewing date; another needs to distinguish an investor enquiry from a request for a rental. Agree the required fields before introducing a score that looks objective but hides assumptions.
A Dubai property brief might include community, buy-or-rent intent, ready-or-off-plan preference, AED budget, timing and preferred contact method. Store what the customer actually said. If a budget is absent, label it as missing instead of inferring it from language, nationality or a social profile.
Keep the original conversation attached
A short AI summary helps an agent prepare, but it should not replace the source. Put the relevant message and its timestamp beside the structured fields so a reviewer can check whether the interpretation is justified. Preserve uncertainty when the message contains conflicting preferences.
For social-media lead intelligence, access and outreach permissions are separate decisions. Finding a relevant comment does not automatically authorise extracting extra personal information or sending unsolicited messages. Use approved channels and have the business review the data flow before connecting it to sales activity.
A lead score is less useful than a clear explanation of what the customer asked for.
Route with rules the team can explain
Once the request is structured, routing often belongs in ordinary business rules. Assign by branch, community coverage, service type or an agreed workload method. Show the rule that assigned the enquiry so managers can correct a poor match without reverse-engineering an AI decision.
Detect duplicates before creating another customer record. If the same person has two active requests, preserve both requirements while linking them to the appropriate contact. A clean ownership model prevents two agents from making conflicting promises to the same customer.
Keep availability and advice outside the model’s imagination
An assistant should not invent a listing, confirm a property is available or promise an investment return. Retrieve current approved information from the source system, and route consequential questions to an agent. A confident response is not evidence that a property record is current.
The first version can prepare a draft response and viewing task. An agent checks the proposed property, price, availability and tone before anything is sent. This gives the team a useful checkpoint without forcing them to rewrite every field manually.
Pilot one community and one channel
Choose a manageable slice, such as inbound enquiries for one community. Assemble a review set containing straightforward leads, vague requests, duplicates and enquiries outside the agency’s scope. Have agents agree what a good brief would contain for each example.
Run the workflow in parallel with the current process. Track missing fields, incorrect classifications and time to a reviewed response. Do not optimise only for the number of leads labelled qualified: an inflated qualification rate can make the sales team’s work harder rather than better.
Turn the pilot into a maintainable sales tool
Before launch, name an owner for the qualification rules, listing source and exception queue. Decide how the system behaves outside working hours and when an agent is unavailable. Include a way to pause automated actions while keeping new enquiries visible.
Fursah is a relevant product example in our portfolio, focused on buying-intent signals in social conversations. That experience helps frame the problem, but it is not a promise of conversion performance for a new agency workflow. Define your own baseline and review the pilot with the agents who will use it.
Common questions.
Put the idea to work.
Prepared with AI assistance. This article shares general guidance and illustrative workflows, not a substitute for advice specific to your business.
Business thinking. Technical doing.
Talk about your workflow ↗
