Customers text a pickup order to a café's WhatsApp number in their own words. An AI agent matches it to the menu, asks about missing options, confirms the total and a pickup slot, and sends the order to the POS and the kitchen display.
Typical results
Under 2 min
From first message to a confirmed order
≈90%
Orders confirmed without staff typing
≈85%
Fewer missed or mistyped orders
Figures reflect typical results for this kind of automation.
The challenge
Morning pickup orders arrive as WhatsApp messages while the counter line is already long. A barista reads each one, types it into the POS and replies to ask about milk or size. Orders get missed or mistyped, some are promised for slots the kitchen cannot fit, and regulars end up waiting for drinks nobody started.
Before and after
Staff time on pickup order-taking per week
−83%Missed or mistyped orders per week
−86%Minutes from message to confirmed order
−80%The automation
An AI agent running in Make with OpenAI reads each WhatsApp order, like two croissants and an oat latte at 8:15, and matches it to the menu. It asks about missing options and replies with the total and a pickup slot the kitchen can fit. Confirmed orders go into the POS and onto the kitchen display with the pickup time, and a ready message goes out when staff tap done. Unmatched messages reach the counter team as flagged chats.
How it works
- 01
Order arrives on WhatsApp
A customer writes an order in plain words, with items and a pickup time, to the café's WhatsApp number.
- 02
Menu matching
The AI agent maps each item to the menu and asks about any missing choice, such as milk type or cup size.
- 03
Total and pickup slot
It replies with the total and the closest slot that still has kitchen capacity, then waits for the customer to confirm.
- 04
POS and kitchen display
The confirmed order is created in the POS and shows on the kitchen display, sorted by pickup time.
- 05
Ready message
When staff tap done, the customer gets a ready message. Chats the agent cannot match are flagged for the counter team.
Starts from
- WhatsApp order messages
- Menu and slot capacity
Delivers
- Order created in the POS
- Kitchen display ticket
- Ready-for-pickup message
The impact
Orders are taken, priced and timed without a barista stopping to type, which saves about 10 hours a week of order-taking. Missed or mistyped orders fall from around 14 a week to 2, pickup slots match what the kitchen can produce, and customers walk in to a bag that is already waiting.
Questions about this automation
How does the agent handle items that are not on the menu?
The agent works from the café's current menu, including sizes, milk options and extras. When a customer asks for something it cannot match, such as an off-menu drink or a large office order, it says a team member will reply and flags the chat for the counter. Sold-out items are switched off in the menu sheet.
Which systems does it connect to?
It runs on the café's WhatsApp Business number, writes orders into the POS so they print or display like counter orders, and reads kitchen capacity per 15-minute slot from a settings sheet. The agent logic runs in Make with OpenAI, so the POS connection can be swapped if the café changes systems.
How does it avoid promising slots the kitchen cannot make?
Each pickup slot has a cap on drinks and food items set by the owner. The agent counts confirmed orders in every slot and offers the next open one when a slot is full. This is a multi-step build, usually rolled out in stages, often starting with drinks before food is added.