Restaurants in Latin America lose 20-301% of reservations due to no-shows and between 40-601% of leads due to not responding promptly on WhatsApp. The solution isn't to buy another app: it's to schedule within the chat with an AI agent, automatic reminders, and optional deposit. Aurora Inbox connects WhatsApp Business, a booking engine, and pre-approved templates so the maître d' can stop carrying their phone and Friday and Saturday occupancy increases by 10-151%.
Why WhatsApp is winning over OpenTable and Resy in Latin America
In Mexico, Colombia, Argentina, Chile, and Peru, WhatsApp penetration exceeds 951% of smartphone users. OpenTable and Resy together have around 25-301% awareness and significantly less actual usage. Latin American diners don't open the OpenTable app because they already have WhatsApp open all day.
Practical differences:
| Variable | OpenTable / Resy | WhatsApp with AI agent |
|---|---|---|
| Customer response rate | 30-40% | 80-95% |
| Customer needs to install something | Yeah | No |
| Confirmation of attendance | Email (15-25% read) | WhatsApp (95% reading) |
| Cost per reservation | $1-2.50 USD per cover | Helpful messages almost free |
| Local language LATAM | Limited | Native |
| It works in towns / tourist areas | Only in Mexico City and Buenos Aires | Where there is 4G |
WhatsApp won because the customer was already there. The only thing missing was the restaurant's five-minute delay in replying.
The typical booking flow via WhatsApp
A diner in Polanco sending a message on a Friday at 6 PM:
Customer: Hi, is there a table for 4 available tonight around 9? Agent: Hi Sofia, sure. For today at 9:00 PM, would you prefer the terrace or the main dining room? We have space in both. Customer: Terrace if possible. Agent: Perfect. Table for 4 on the terrace, today, April 30th, at 9:00 PM. What name should I put it under? Customer: Sofia Ramirez. Agent: Done, Sofia. Reservation confirmed: - 4 people, terrace - Today at 9:00 PM - We'll hold the table 15 minutes later. Location: maps.app.goo.gl/xxx Menu: aurorainbox.com/menu.pdf We'll remind you tomorrow... I mean, I'll confirm 2 hours before. If you need to cancel, reply CANCEL right here. [8:00 PM — automatic reminder] Agent: Hi Sofia, we'll be waiting for you tonight at 9:00 PM on the terrace, 4 people. Do you confirm your attendance? Please reply YES or CHANGE. Customer: YES. Agent: Great. See you in a bit.
Five to seven turns in total. The maître d' didn't respond. The table was blocked with a 2-hour turnaround buffer.
Measurable benefits before vs after
| Metrics | No system (phone + notebook) | With WhatsApp + Aurora Inbox |
|---|---|---|
| Response rate to incoming messages | 40-60% | 95-100% |
| Lead rate to confirmed booking | 25-35% | 60-75% |
| No-show rate | 22-30% | 10-15% (5-7% with tank) |
| Average weekend occupancy | 70-80% | 85-95% |
| Host time on phone | 3-5 hrs/day | 0-30 min/day |
| Average ticket price (with pre-order upsell) | Baseline | +8-15% |
| Sales attributable to reminders | $0 | 5-12% of revenue |
The figures vary by vertical (casual, fine dining, hotel) but the direction is the same for every client we onboard.
Aurora setup: how to model a restaurant
The model we use for clinics and salons is seamlessly mapped to the restaurant. Three concepts:
Services = types of tables
- "Table for 2 people — 90 min"
- "Table for 4 people — 120 min"
- "Table for 6 or more people — 150 min"
- "Private table — 180 min"
- "Group meal (10+) — 240 min, deposit required"
The duration determines the availability block. If you set 90 minutes for a 2-player table, the engine opens slots every 90 minutes at that table. A 150-minute block for a 6+ player table better reflects real-world behavior (long dinners).
Location = the restaurant (or branch)
- Name, address, and time zone. It's critical that the time zone is that of the location, not the customer's — a Marriott guest booking from Madrid needs to see Mexico City time, not Madrid time.
Service providers = zones, not waiters
Here's the key insight. Don't model a server as a service provider. Model zones or rooms:
- "Terrace — capacity 6 tables simultaneously"
- "Main Hall — 14 tables"
- "Bar — 8 places"
- "Private — 1 table of 12"
Each zone has its own hours and simultaneous capacity. The system allows multiple simultaneous reservations within each zone's capacity. If you want to go down to the level of individual tables (typical in fine dining), you can—one server per numbered table. For casual dining, this isn't worthwhile; it would over-engineer the operation.
Aurora supports simultaneous capacity per provider, meaning that the "Main Hall" with a capacity of 14 accepts 14 reservations in the same slot. Step-by-step guide at Configure services, locations, and providers and the agent's complete flow in How to set up a WhatsApp appointment system with Aurora Inbox.
Typical buffer between reservations at the same table: 30 min. This prevents the 7:00 pm customer from crossing paths with the 9:00 pm customer if the former stays until 8:50 pm.
Capabilities that only make sense in restaurants
Dynamic waiting list. When all slots are booked, the agent offers to be placed on the waitlist. If someone cancels, the agent automatically notifies the first person on the list and holds the reservation for 10 minutes for confirmation. This recovers 5-8% of reservations that would otherwise be lost.
Minimum number of people at large tables. The "Private Table" or "Group Meal" service rejects reservations for fewer than 6 people (or the number you configure). The agent will redirect you to the appropriate service.
Pre-order via WhatsApp. The customer can order appetizers, desserts, or drinks before arriving. This reduces time spent at the table, increases the average check, and improves turnover. Similar pattern to Recover abandoned e-commerce shopping carts via WhatsApp — a link is sent and the client confirms.
Reminder with menu and location. Pre-approved template including a PDF menu and Google Maps link. Upload the ticket because the customer arrives with an idea of what they want to order.
Mandatory confirmation 24 hours in advance. The template sends the message "Can you confirm your reservation for tomorrow?". If there's no response within 6 hours, it automatically cancels the reservation and releases the table to the waitlist. Why it works: explained in What is no-show and how to reduce it and Reduce no-shows with WhatsApp reminders.
Special case: fine dining with deposit
Restaurants with an average check over 1,500 MXN (1,500 USD) per person should charge a deposit. The flow:
- Customer requests reservation for 4 people on a Saturday.
- Agent: "To confirm, we require a deposit of 1,430 MXN per person, refundable if you cancel 24 hours in advance. Here's the link."
- Aurora generates Stripe (or Conekta / Mercado Pago) link for $1,200 MXN.
- If you pay within 30 minutes: your booking will be automatically confirmed.
- If you do not pay: the reservation expires and the slot is released to the next person.
- If he arrives and fulfills the requirement: the deposit is deducted from the account. If he doesn't arrive: he loses it.
Typical result: no-shows drop from 22% to 4-6% in fine dining. The friction of the deposit filteres out the serious diner from those who book just for the sake of booking.
Real case: casual restaurant in Mexico City
One of our clients with 80 tables in Roma Norte, ~600 reservations per week, average ticket $650 MXN.
Before (Q4 2025). A host with a landline, paper notebook, and personal WhatsApp. 28% no-shows on Fridays. Average response time to an incoming message: 22 minutes. Lead-to-booking: 31%.
90 days after implementing Aurora Inbox.
- AI agent answers in less than 30 seconds, 24/7.
- Reminder 24h and 2h before with mandatory confirmation.
- 5 zones modeled as providers with real capacity.
- No-show: 28% → 12%.
- Lead-to-reserve: 31% → 64%.
- Occupancy Friday and Saturday: +15%.
- Host now works in the office — not by phone — and serves in-person customers who previously waited 5 minutes without anyone greeting them.
- Recovery with waitlist: 27 extra bookings per month.
Calculated ROI: Monthly subscription represents 0.4% of the extra revenue recovered.
Common mistakes (the six we see repeated)
- Treat each table as a provider. It ends with 80 providers and an unmanageable operation. It models zones; only go down to the table if you do premium fine dining.
- Do not configure actual simultaneous capacity per zone. If the terrace has 6 tables and you left a capacity of 1, the engine only allows one reservation per slot.
- Reminder without cancellation link. The customer doesn't know how to cancel, doesn't notify you, and becomes a no-show. Always include "reply CANCEL" or a direct link.
- Agent confirms reservation without verifying availability. This happens when the bookings plugin isn't active and the agent improvises. Always activate the BookingPlugin on the agent's device—the engine checks actual availability before confirming.
- Not measuring anything. Without a dashboard tracking no-shows, occupancy, and lead-to-booking, you won't know if the change worked. Aurora offers built-in metrics; review them weekly.
- Ignore the waiting list. It's the easiest ROI to capture. Activating it takes 2 minutes and recovers 5-8% of revenue every week.
More context on the conversational pattern in What is conversational scheduling?.
How does Aurora Inbox do it?
Aurora Inbox combines WhatsApp Business Cloud API at the BSP level, an AI agent with a booking plugin (simultaneous capacity engine, rotation buffer, waitlist), automated reminders with pre-approved Meta templates, and optional deposit via Stripe/Conekta/Mercado Pago. For restaurants in Latin America, it's the most direct way to reduce no-show fees below 15% without taking the customer out of WhatsApp.
Recommended plan for a restaurant with ~500 reservations/month: Aurora IA $179 USD/month ($3,200 MXN) — Includes AI agent with booking plugin, reminders, and metrics dashboard. For hotel groups or chains with multiple branches: Aurora IA Plus $329 USD/month ($6,000 MXN)If you only want a human team without AI: Aurora CRM $99 USD/month ($1,800 MXN).
Start your free trial — the complete setup of a restaurant takes 30-45 min.
Frequently Asked Questions
Does Aurora Inbox integrate with POS Toast or Square?
Yes, via REST API and webhooks. Toast and Square expose endpoints to insert the reservation as a "guest profile" and bring the checkout into the system. Custom integration typically takes 1-2 weeks with your team or partner. Mercado Pago Point and SoftRestaurant also support webhooks.
Is WhatsApp Cloud API the best option for restaurants?
Yes. It's the official Meta API, offering better message quality and faster template approval. Aurora Inbox uses the Cloud API directly at the BSP level, eliminating intermediaries and reducing the cost per message to as low as 30%.
Is it legal to accept card deposits via WhatsApp?
Yes, in Mexico, Colombia, Argentina, Chile, and Peru, provided the payment is processed through a certified gateway (Stripe, Conekta, Mercado Pago). Aurora does not store card details—the customer pays through the gateway, and the restaurant receives the confirmation. Check with your accountant regarding the tax implications of the deposit (it's generally treated as an advance payment).
Does it work for chains with multiple branches?
Yes. Each branch is modeled as a distinct location with its own zones (providers) and hours. A single Aurora workspace can handle 1 to 50+ branches. At the start of the chat, the customer indicates which branch, and the agent checks availability only at that branch.
How do you handle bookings for large groups (more than 20 people)?
Create a "Private Event / Group 20+" service with a long duration (4-6 hours) and configure the agent to automatically escalate to a human when someone requests that service. These cases typically require a closed menu, a larger deposit, and a conversation with the chef or manager—not ideal for 100% AI.
What about walk-ins (without a reservation)?
Aurora handles walk-ins as manual check-ins from the calendar view: the host registers the occupied table, and the system deducts capacity from that area in that slot. For a more sophisticated flow, you can generate a QR code at the table that opens WhatsApp, allowing the customer to "check in"—useful for tracking time spent at the table and closing the digital check.
