CUSTOMER STORYFOOD · HEALTHY DESSERTS · ONLINE STORE & PICKUP
The sweetestnight shift.
How a healthy-dessert patisserie answers every customer on WhatsApp in seconds, from “will the chocolate melt in delivery?” to “is it OK for diabetics?”, even at 1am and on Shabbat.
- 84%
- of after-hours messages answered within 5 minutes
- 12 sec
- median time for the AI agent to answer
- 81%
- of replies to automatic follow-ups answered first by the AI
- +74%
- more customers writing each month since July
Measured on the patisserie’s live AssistantLabs account, 2026. Business name withheld.
THE STORY
Healthy desserts raisevery personal questions.
The business
A patisserie that makes desserts with no added sugar, rich in protein and fibre: brownies, pralines, bars, spreads and gift boxes. It sells through its own online store, with delivery and self-pickup, and gets enquiries from cafés and businesses too.
The challenge
Customers don’t just ask about price. They ask whether a dessert suits diabetes or pregnancy, whether it’s gluten-free or kosher, whether chocolate survives delivery in summer, and when they can pick up. They ask in the evening and on weekends, when they’re planning a treat or a gift, and many start an order and don’t finish it.
What AssistantLabs connected
- An AI agent on WhatsApp that knows the menu, the ingredients and the store’s policies: 8 sets of FAQs, the product catalog and 177 pages from the website.
- Flashy, the store’s marketing automation: when a shopper starts checkout and doesn’t finish, a WhatsApp message goes out, and the agent picks up the reply.
- Business leads: cafés, events and corporate orders are qualified and handed to the owner as leads.
- Labels for the team: pickup, delivery, sales, club and more, so everyone sees what each conversation is about.
- Handoff with context for anything that needs a person, with the whole conversation attached.
- A shopper leaves checkoutor writes on WhatsApp, at any hour
- Flashy sends a WhatsApp“Did something happen? Maybe I can help?”
- The AI agent answersingredients, diets, delivery, pickup, business enquiries
- The team steps infor pickups, special orders and business leads, with the context
AFTER HOURS
At night, the answerscome even faster.
During the day the team answers many messages itself. After closing, the AI agent takes over, and customers wait less, not more.
Thursday · 1:32 am
- 68%
- of after-hours messages answered first by the AI agentvs 41% during opening hours
- 1 in 4
- customer messages arrives after closing2,383 in six months
- 9%
- of conversations ask about sugar, diabetes or ketokeyword-based estimate
Answered within 5 minutes
Share of customer messages, by when they arrived · number of messages
Real conversation, translated from Hebrew and lightly shortened. Figures: 27 Mar – 25 Sep 2026. After hours includes night and Shabbat. Shabbat: Fri 4pm – Sat 8pm.
UNFINISHED ORDERS
“Did something happen?”Then the agent takes it from there.
When a shopper starts checkout and leaves, the store’s marketing system sends a friendly WhatsApp message. When the customer answers, with a question about pickup, delivery or ingredients, the AI agent is already there.
Tuesday
- 20%
- of unfinished-checkout follow-ups got a reply within a day1,949 follow-up messages in 2026
- 85%
- of those replies answered first by the AI agentin a median of 16 seconds
- 3,851
- automatic follow-ups sent through Flashy in 2026the unfinished-checkout message plus cart reminders
- 1,276
- shoppers who left checkout got a personal follow-upsent on their own, with no one on the team pressing send
Real conversation, translated from Hebrew and lightly shortened. Figures: 1 Jan – 26 Sep 2026. “Wrote back” = the customer replied within 24 hours.
BUSINESS CUSTOMERS
A café chain asks at 11:21.A tasting is booked the same day.
Sunday
Later that day: a tasting meeting booked for Wednesday
- 12/13
- business enquiries answered first by the AI agentcafés, a festival, brand collaborations, corporate orders
- 14 s
- median first answer to a business enquiry5 of them arrived after hours
- 12
- logged as leads and passed to the ownerafter asking for quantities, frequency, packaging and contact details
What customers ask about
Share of conversations, keyword-based estimate · topics overlap
Real conversation, translated from Hebrew and shortened; names and business details removed. Business enquiries found by keyword search and reviewed by hand, 2026, so the count is a floor.
THE RESULTS
More customers every month,more of them handled end to end.
Customers who wrote on WhatsApp each month, 2026
Customers writing that month% = handled by the AI agent with no one from the team
*September: 1–26 Sep. From February the share handled by the agent alone climbed from 18% to 32%, while the number of customers grew.
I was really impressed by the AI assistant you have on WhatsApp. It’s simply amazing!
Want this for your business?
In a free integration meeting we sit down with you on your business and decide what no longer has to go through you. For example:
- Every enquiry answered in seconds, even at 1am
- Unfinished orders followed up on WhatsApp, automatically
- Business leads qualified before you pick up the phone
- Every conversation labelled, for the whole team
About the data
Measured on the patisserie’s live AssistantLabs account. Speed and topic figures: 27 Mar – 25 Sep 2026; monthly chart and follow-ups: 1 Jan – 26 Sep 2026. A customer message is one or more lines sent together; reply time runs from that message to the first reply. After hours: outside Sun–Thu 9am–6pm and Fri 9am–1pm, Israel time. “+74%”: average customers per month in Jul–Aug (350) vs Jan–Jun (201). Topic shares are keyword-based estimates.
More customer stories
- Custom car matsFrom first question to front door.~700 customers write in on WhatsApp every monthRead the story ›
- Craft storeThe craft store that never closes.40% of customer messages arrive outside business hoursRead the story ›
- Lighting storeEvery phone call continues on WhatsApp.63% of new callers continue the conversation on WhatsAppRead the story ›
