Case study · Cleaning services · Dubai
Amira answers every call, quotes the job and books it — in English or Arabic
A Dubai cleaning company was losing enquiries after hours and quoting inconsistently during them. We built a voice agent that handles the entire inbound sales conversation end to end, and put it into production.
The problem
Calls outside hours went to voicemail, and quotes varied by whoever answered.
What we built
A bilingual voice agent that qualifies, quotes from the real rate card and books.
Integrated with
Scheduling calendar, CRM and WhatsApp confirmations.
Status
Live in production, handling real customer calls.
- Caller
Hi, do you do deep cleaning for a three-bedroom villa?
- Amira
We do. For a three-bedroom villa a deep clean is usually a team of three for around five hours. Is the villa furnished and occupied at the moment?
- Caller
Yes, we live there. How much roughly?
- Amira
For that size and condition it comes to 780 dirhams including materials. I can hold Thursday morning or Saturday afternoon this week — which suits you?
- Caller
Saturday afternoon please.
- Amira
Booked for Saturday at 2pm. I have sent the confirmation to your WhatsApp with the team lead’s name. Anything else I can help with?
An excerpt from a real call. Ask us for the full recording on a readiness call.
The situation
A good business quietly leaking work
The client is a well-run cleaning company with a strong reputation and a busy inbound line. Their problem was not lead generation — people were calling. The problem was what happened to those calls.
Enquiries arriving in the evening or at the weekend hit voicemail, and by Monday a proportion of those callers had already booked with someone else. During working hours the line was answered, but the quote depended on who picked up. Two callers with identical villas could be told two different numbers, which caused friction later when the team arrived on site.
Roughly half their market prefers to be dealt with in Arabic, which meant the quality of a caller's experience also depended on who happened to be on shift.
How we built it
Four weeks, and none of it started with a model
-
We listened to the calls first
Before writing anything we went through real recordings of their inbound line. Most callers asked the same four things in a different order, and the agents were improvising the price each time. That inconsistency, not the call volume, was the actual problem.
-
We modelled their pricing, not a generic script
Amira quotes from the client’s real rate card: property size, occupancy, materials, team size and duration. She does the arithmetic the way their best estimator does it, so the quote a caller hears is the quote the business honours.
-
We connected her to the systems that matter
Availability comes from the live scheduling calendar, not a static list. When she books, the job is written to the calendar and the CRM, and the confirmation goes out over WhatsApp automatically.
-
We tested her against the team before any customer
The operations team tried to break her for two weeks — odd accents, code-switching mid-sentence between English and Arabic, callers changing their mind, edge-case properties. Every failure became a flow fix before go-live.
What it does now
Every call answered, every quote the same
No more voicemail
Calls at any hour are answered on the first ring, qualified and either booked or passed to a person with the context attached.
One consistent price
Quotes come from the client's own rate card every time, so what the caller is told is what the business charges.
Served in their language
English and Arabic handled equally well, including callers who switch between the two mid-sentence.
We publish what we can evidence. Commercial figures for this engagement are the client's to share, not ours — we will walk you through the detail on a call.
What would this look like for your business?
A dealership, a clinic and a hotel need very different agents. The first call is us working out which conversations in your business are worth handing over.