How to create a WhatsApp AI agent for your coffee shop

Wabery lets you set up a WhatsApp AI agent that answers your coffee shop’s questions and reserves tables using configured availability. You can manage it yourself in the dashboard without code. This tutorial covers menu answers and table bookings; takeaway ordering, payment collection, and live food stock need a separate setup.

What can a WhatsApp AI agent do for a coffee shop?

A Wabery agent can answer questions from your saved business information, check table availability, and create a reservation after the customer confirms. Your staff can take over in Inbox. What it knows depends on the information and booking rules you supply.

If you only need opening hours and menu answers, start with Customer support. Use Reservations for the table-booking walkthrough below.

How do I set up a coffee shop agent without coding?

Open Wabery’s agent setup, choose Reservations, and describe the job. Wabery hosts the agent; you supply its information and review its behavior. You do not need an agency, although someone can set it up for you.

1. Describe what the agent should answer and when staff should take over.
2. Choose its voice and reply language preference.
3. Paste your business information, import your website, or upload a text, Markdown, or CSV file. Review imported facts.
4. Enable Smart Replies for answers using that knowledge, then select Build the agent.

Here is an example instruction you can adapt:

Answer questions about our café’s menu, opening hours, and tables. Keep replies short. For reservations, ask for the date, time, and number of guests. Check availability and ask the customer to confirm the exact details before booking. Never invent prices, stock, or availability. Ask the team about complaints and information you cannot verify.

See the hosted-agent guide for the full setup controls.

What information should I give the agent?

Give it the facts you would give a new member of staff: your menu, prices, address, opening hours, facilities, and customer policies. Instructions such as “be helpful” cannot replace those facts.

For a fictional café, your starting note could be:

Maple & Bean: 12 Market Lane. Open Monday–Saturday, 8 am–5 pm; Sunday, 9 am–2 pm. Latte $4; oat milk add-on $0.50. Wi-Fi available. Dogs welcome on the terrace. Reservations last 60 minutes. Parties over four need staff approval. Customers order and pay at the counter.

Replace every example with your own details. Add cancellation rules, holiday closures, and verified ingredient information. Tell the agent to ask staff about allergies or cross-contact rather than promise an item is allergen-free.

How do I let customers reserve tables on WhatsApp?

Configure Tables in Wabery Bookings, publish the booking rules, and launch agent booking actions. Putting table details in the agent’s knowledge alone does not create bookable inventory.

1. Open Bookings → Booking settings and choose Tables. Each project has one venue and one reservation model; the model is fixed after the first successful save.
2. Set the venue timezone and add actual tables, such as two tables for two people and one for four.
3. Set bookable hours, reservation duration, cleanup time, closures, and customer policies. Only configure table combinations staff can actually arrange.
4. Save, inspect the availability preview, turn on Bookings, and select Publish booking changes.
5. Connect your number through Channels → Connect, then review and publish your agent in Agent.
6. Complete the booking readiness checks and select Launch agent booking actions.

These are separate controls: connecting WhatsApp does not publish your agent, and publishing booking rules does not launch customer-chat bookings. See Reservations.

How does Wabery avoid booking the same table twice?

Wabery checks availability again when the customer confirms and allocates the table only after that confirmation. Your team must also record phone bookings and walk-ins in Bookings → Calendar. A reservation written only in another calendar is not automatically blocked in Wabery.

A booking conversation should propose an exact date, time, party size, and duration before asking for confirmation. Check the saved calendar record rather than relying only on the reply text.

Can the agent take takeaway orders or payments?

The setup in this tutorial cannot place takeaway orders, collect payment, or check live kitchen stock. A saved menu lets the agent explain items and prices. Wabery Reservations manages table availability, not food orders or checkout. Keep ordering with your staff or use a separately configured ordering integration.

Can I keep my existing WhatsApp Business number?

Wabery has a Keep using the Business app connection path, subject to the setup requirements. If another API provider manages your number, choose Move an existing API number instead. Do not delete or deregister the number first. Follow Connect WhatsApp for your situation.

What happens when a customer needs a person?

Configure a keyword such as “human” to run the Pause Replies automation action. A staff member replying manually in Inbox also pauses automation for that conversation. Handoff stays on until someone turns it off; staff must monitor those conversations. See human handoff.

If Smart Replies is off, unmatched messages use your configured fallback automation or go to the team. Set a fallback that tells customers when a person needs to respond.

How much does a Wabery WhatsApp agent cost?

Wabery’s Free plan includes one Live project, one connected channel, and 1,000 messages a month. Pro is $15 a month with two Live projects and 50,000 shared messages. AI usage and WhatsApp delivery are separate from the platform allowance. See current pricing.

AI replies, including private tests, use Wabery Wallet credit. New accounts receive $1 of AI credit without a card. Your first Wallet top-up of $10 or more adds $2 bonus credit. Enabling Smart Replies does not itself add a card, charge a top-up, or enable automatic top-ups.

Meta bills dedicated-number WhatsApp charges separately. Shared Sandbox delivery is free through September 30, 2026; from October 1 it uses Wallet credit at Meta’s published service rate without markup. See WhatsApp pricing.

How do I check the agent before customers use it?

Test the editable Draft in the simulator. You can share a saved Draft through a Private Wabery link, which does not expose live business actions, or use the shared WhatsApp Sandbox from your phone.

Use the café example above to check:

• “How much is a latte with oat milk?” The answer should be $4.50.
• “Do you have gluten-free cake?” It should not invent an item.
• “Can I book for eight people?” It should follow the staff-approval rule.
• “Human, please.” Your configured handoff should pause replies.

After launching booking actions, make one controlled reservation. Verify its details in Bookings → Calendar, then cancel it. Also try an unavailable time. If a confirmation message fails to deliver, review the booking itself: failed notification delivery does not undo a saved reservation.

For other workflows, see the yoga studio tutorial and barbershop tutorial.