Results after four months
38%fewer no-shows
62%of bookings made outside office hours
19hfront-desk time returned each week
3 wksfrom first call to live
Home / Case studies
Three builds, and what actually changed
Every figure below was measured against the three months before launch, and every client has read and approved their write-up.
Dental · 4 clinics · Kuwait
A booking agent that works the night shift
Meridian Dental Group was sending every after-hours caller to voicemail — and watching a third of them book somewhere else by morning.
The problem
Four clinics, one shared reception team, and a phone line that stopped being answered at 5pm. Patients who called in the evening got a voicemail greeting; most didn't leave a message. The team could see the missed calls the next morning but rarely had time to work through them.
No-shows were the second problem. Reminders went out when someone remembered to send them, which was not every day.
What we built
- A WhatsApp agent that reads the live clinic calendar and books, moves or cancels appointments
- Bilingual Arabic and English, chosen by the patient in the first message
- Automatic reminders the day before, with a one-tap confirm or reschedule
- Anything clinical or unusual handed straight to reception with the full thread
- A daily summary so the practice manager can see what the agent did overnight
We were losing patients to voicemail after five. Elitra had a booking agent live in three weeks, and it now books more appointments overnight than our front desk does at lunch.
Freight forwarding · 22 staff
Six hours of quoting, done in eleven minutes
Halcyon Logistics priced every enquiry by hand across three systems, and lost the slow ones to faster competitors.
The problem
A quote meant opening the rate sheet, checking the customer's history in the CRM, looking up the current surcharge, then writing the email. Six hours on average, longer if the person who knew the route was away. Customers who wanted a number the same day often went elsewhere.
What we built
- An agent that reads inbound enquiry emails and extracts route, weight and dates
- Live pricing against the rate sheet, including the surcharges nobody remembered
- A drafted quote in the company template, waiting in the sender's outbox
- One human click to send — pricing judgement stayed with the team
- Automatic follow-up on day three and day seven if the quote goes quiet
Results after six months
11 minaverage quote turnaround, from 6 hours
1,400manual entries removed each month
+31%more quotes sent per week
0quotes left without a follow-up
I expected a demo and a big bill. What I got was a working quote pipeline, my team trained on it, and a number I could take to the board.
Home services · HVAC and plumbing
The calls they used to miss on a roof
Verda Home Services had six engineers in the field and nobody free to answer the phone. Roughly one in three calls went unanswered.
The problem
Every engineer was also the person who answered the phone, which meant they answered it from a ladder or not at all. Emergency callers who couldn't get through called the next company on the list. The founder estimated the loss at tens of thousands a year but had no way to see it.
What we built
- A voice agent that answers on the second ring, day or night
- Triage: genuine emergencies escalate to the on-call engineer's mobile immediately
- Everything else gets booked into the right engineer's diary by postcode and skill
- The job, the address and the notes land in the CRM before the call ends
- A weekly report showing calls answered, jobs booked and revenue attached
We also talked them out of two things they'd asked for: an AI that would quote jobs sight unseen, and a chatbot on a website almost nobody visited.
Results in the first quarter
94%of calls answered, up from 68%
$47krecovered from previously missed calls
4.8customer satisfaction, unchanged after launch
2 minaverage emergency escalation time
They talked me out of two of the three things I wanted to build. That is the moment I decided to keep working with them.
Your business, next
Bring the process that costs you the most time. In 45 minutes you'll have an honest read on whether it's worth automating and roughly what it would cost.