AI Case Studies from Puerto Rico: Real Deployments, Real Numbers

Case 1 — Casa Chunan, Maunabo (Vacation Rental Concierge)

Property: luxury 3-bedroom vacation rental in Maunabo, southeast Puerto Rico. Host: Kimlee Gely, Airbnb Superhost. Guests: mix of mainland US travelers and Puerto Rican locals, split roughly 60/40. Problem: host was fielding guest questions (WiFi password, checkout time, restaurant recommendations, directions) across WhatsApp, SMS, and Airbnb messages at all hours. Response time varied. Some guest inquiries slipped. Review rating was 4.8 and stuck there. Solution: deployed “Lota,” a bilingual AI concierge on WhatsApp + SMS. Trained on the property specifics, local recommendations, and house rules. Language detection per message. Stack: Claude Sonnet 4.6 + Twilio WhatsApp + ElevenLabs Scribe for voice notes. Integrated with Airbnb API for reservation context. Results: host intervention on routine questions dropped ~85%. Guest response time dropped from variable to under 30 seconds. Review rating moved from 4.8 to 5.0 Superhost. Weekend response rate (when the host is offline) went from ~60% to 100%.

Case 2 — Club Brava, San Juan (VIP Nightlife Concierge)

Venue: one of San Juan's premier nightclubs. Owner: Shimmy McHugh. Peak volume: Friday-Saturday nights, event nights, high-profile international DJs. Problem: team was fielding 200+ inquiries per busy night across phone, Instagram DMs, and WhatsApp. VIP table inquiries, guest list additions, event questions. Peak-hour response times slid into hours. Bookings dropped off during the windows they mattered most. Solution: deployed “Marco,” a bilingual VIP concierge on phone, Instagram DMs, and WhatsApp. Quotes bottle service, builds guest lists, RSVPs events, answers event-specific questions — all while the team focuses on in-venue operations. Stack: Retell AI for phone routing + Claude for reasoning + ElevenLabs voice + Meta Business Suite for Instagram/WhatsApp. Voice is a native Puerto Rican speaker clone. Results: inquiries answered instantly at any hour. Bottle-service quote lead time collapsed from hours to seconds. Peak-night booking capture up meaningfully. Team reports a clean separation between in-venue focus and inbound handling for the first time.

Case 3 — Bayamón Auto Repair Shop

Business: independent auto repair shop in Bayamón. Owner answers 50+ calls per day during business hours plus overflow on his cell after hours. Problem: owner was the bottleneck. Every service call interrupted him in the shop. Missed calls meant missed customers — Bayamón has enough competition that a customer who didn't get through once rarely called back. Solution: bilingual AI answers every incoming call, quotes parts + labor from a pricing sheet, and books the service appointment directly into the shop's Google Calendar. Stack: Twilio voice + Claude + Google Calendar API. Pricing logic hard-coded from the shop's standard rate sheet. Results: owner-side interruptions down from 50+ calls/day to 5–10 (only the ones that needed him). Bookings up 22% in the first 90 days. A non-obvious second-order effect: with fewer interruptions, the shop's labor hours per vehicle dropped — meaning each car actually got out the door faster.

The Three Lessons We Took Forward

WhatsApp beats phone in PR whenever the choice exists. For every deployment that touched both channels, WhatsApp volume meaningfully exceeded phone volume within 30 days of launch. Deploy WhatsApp first if you're picking one.
Second-order effects matter more than the primary metric. In the auto shop case, fewer interruptions didn't just save the owner time — it made his labor hours more productive. Look for these compounding effects when sizing ROI.
Bilingual is not a feature, it's the floor. Every single PR deployment had bilingual customers from day one. Plan for it at the model layer, not as an add-on.

What This Means if You're Considering AI

If you run a hospitality, service, or local-retail business in Puerto Rico and you handle more than ~30 customer interactions per day across any combination of phone, WhatsApp, Instagram, or web chat — AI automation probably pays for itself in month 1. The cases above aren't edge deployments. They're representative of a typical month of work for us on the island. What differs from one business to the next is the stack and the integration surface — not the underlying ROI math.

Frequently Asked Questions

Can I reference any of these Puerto Rico AI deployments publicly?

Casa Chunan (casachunan.com) and Club Brava are public. Direct case study materials are available on request during discovery calls.

Does Dreamer AI work with non-hospitality businesses in Puerto Rico?

Yes. The case studies above lean hospitality and auto because those are highest-volume verticals. We deploy for clinics, legal offices, real estate, home services, and retail — the underlying pattern (bilingual + WhatsApp + AI) applies broadly.

What's the minimum volume to justify AI for a PR small business?

Rough threshold: ~30 customer interactions per day across all channels. Below that, the ROI case gets thin. Above it, AI usually pays for itself in month 1.

Are these numbers typical or best-case?

Representative. These are the median outcomes for our PR deployments — neither the top quartile nor the bottom. Outcomes vary by business readiness, data quality, and integration surface.

How do I get a custom ROI estimate for my PR business?

Book a free 30-minute discovery call. We map your workflow, identify the highest-leverage leak, and quote specific monthly and first-year ROI ranges for your situation.