
Illustrative scenario
This is a written composite, not a real customer. The problem, the configuration and the numbers are the kind we expect to see when minidesk runs in production — they are not measurements from a named business. The first real, on-the-record case study lands the day a customer agrees to be quoted.
How a PJ bubble tea chain cut customer WhatsApp reply time from 4 hours to 2 minutes
F&B, three outlets, custom-order WhatsApp on a same-minute clock
Headline
2 min
WhatsApp reply time
Powered by
StoreHub POS · WhatsApp Business API
The problem
Three outlets, one WhatsApp number, and one person — the owner's daughter — answering every customer message between 11 am and 10 pm. Customisation questions (less sugar, less ice, oat milk, brown sugar, no boba), order changes, pickup times, delivery queries, the rare complaint. The phone buzzed 100–150 times a day. Average reply time was four hours during peak, and the customers who did not get a reply in the first few minutes ordered from the competitor down the road.
What the AI does today
minidesk sits between the StoreHub POS and the WhatsApp inbox. The AI reads the menu, the day's stock levels, the day's running sales, and the order history. It replies to customisation questions, confirms orders against today's stock, gives pickup ETAs from the kitchen screen, and only escalates to a human when the customer asks to speak to a person or when the AI is below its confidence threshold.
The numbers
Average reply time
2 min
was 4 hours at peak
Of messages handled without a human
82%
Customer complaints about slow replies
Down 71%
was month over month
What the owner says
“My daughter used to spend seven hours a day on WhatsApp. Now she spends it on the floor of the busiest outlet — which is the job she actually wanted, and the place the customers come to see her.”
— Mei Ling, Owner, three-outlet bubble tea chain in Petaling Jaya