A WhatsApp Business bot for a CDMX restaurant with a dynamic PDF menu sends the updated menu in seconds, takes reservations connected to your real calendar, collects deposits via Stripe, and logs every customer into your own CRM. Total investment starts at $4,500 USD one-time, operational from $200 USD/month. Your bot handles WhatsApp 24/7 in your restaurant's voice — customers won't notice the difference. No retainers. Code is yours.
If you run a restaurant in Polanco, Roma, Condesa, Coyoacán, San Ángel, or anywhere in CDMX and WhatsApp orders are getting out of hand, this post shows you how to build it.
Why a Dynamic PDF and Not a Static Menu
The average CDMX restaurant changes its menu 3 to 5 times a year: spring, summer, fall, winter, plus special seasons. If the menu is static and lives in a PDF uploaded to Drive, this is what happens:
- Regular customers receive an outdated link (dishes you no longer serve)
- Staff manually sends PDFs over chat every time (40 sends per day = 4 hours of lost labor)
- Weekend promotions don't propagate
- There's no way to know which dish gets asked about most on WhatsApp
A dynamic PDF lives in your own storage. Update it once, and every customer receives the current version. The bot sends the link, logs which customer requested which menu, and leaves the full trail in your CRM.
The Real Case: 26.5% Conversion, 7,197 Organic Sessions
Although the documented case comes from an educational institution, the mechanics are identical for hospitality. Real metrics from the case:
- 113 total conversations handled by the bot
- 30 reservations closed (26.5% conversion)
- 79 automated follow-ups at 24, 72, and 168 hours
- 57 handoffs to a human when complexity was detected
- 7,197 organic sessions in 60 days with zero additional ad spend
- 32.9% bot conversion vs. 14.1% paid traffic conversion
For a restaurant, those metrics translate to: confirmed reservation rate, recovered no-show rate, average ticket per returning customer.
Technical Stack for a CDMX Restaurant
What we recommend for 1 to 3 locations:
| Component | Tool | Function |
|---|---|---|
| Channel | WhatsApp Business API via Twilio | Messaging |
| PDF Menu | Supabase Storage with signed URLs | Current version |
| Calendar | Google Calendar API or proprietary system | Reservations |
| Payments | Stripe Mexico | Deposits and charges |
| Bot | Python backend + Claude or GPT-4 | Responses |
| CRM | Supabase Postgres with returning customers | Pipeline |
Deep integrations with your existing reservation system (Resy, OpenTable if applicable) are evaluated during discovery. We do not use turnkey bot platforms.
Typical Conversational Flow
Saturday night reservation use case:
- Customer: "I want a table for 4 on Saturday at 9pm"
- Bot: "Hi Sofía, let me check. We have 8:30 or 9:15 available. Which works for you?"
- Customer: "9:15"
- Bot: "Confirmed. I'll send you a reminder Saturday at 7pm. Here's the updated menu: [PDF link]"
- Bot sends reminder 2 hours before
- Customer confirms, record is in the system
Total time: 60 to 90 seconds. Zero phone calls. Zero lost reservations from missed responses.
Handling Seasonal Menus and Promotions
The PDF menu is updated once in storage. The bot:
- Sends the current menu link to every new customer
- Quotes dishes when the customer asks without opening the PDF (prices queried from the database)
- Suggests pairings based on a returning customer's history
- Proactively mentions the daily special without being asked
Promotions are loaded into a separate table. The bot decides when to mention them based on context — not based on a rigid template.
Cost and Timeline
MAGIA / Solo covers the use case for an independent restaurant or small restaurant group:
- $4,500 USD one-time
- 15 calendar days
- WhatsApp bot + dynamic PDF menu + reservations + CRM + Stripe
- 3 training sessions for the manager or operator
- 4 weeks of support
Operating cost: $200 to $400 USD/month pass-through. No monthly licenses. No conversation caps.
Common Mistakes from Restaurants That Already Tried
Three recurring patterns:
- They contracted a reservation SaaS that charges per table and doesn't export customer history
- They deployed a generic bot with no training on their actual menu or restaurant voice
- They still send the PDF manually over chat (4 hours of wasted operations per day)
All three are avoided with a proprietary architecture. Your customers are yours. Your menu is yours. Your bot is yours.
Next Steps for Your Restaurant
If your CDMX restaurant handles more than 30 weekly reservations via WhatsApp or more than 100 menu inquiries, the bot pays for itself within 2 months in operational time savings alone.
- MAGIA / Solo for an independent restaurant or 1 to 3 locations
- MAGIA / Core for a restaurant group with 4 to 30 locations and fragmented operations
30-minute call with the team that actually builds it — not with an SDR.