We run a private driver service in Marrakech, Morocco. Five vehicles, one office, no venture capital. When people talk about AI in travel, they mean billion-dollar platforms with recommendation engines trained on millions of bookings.
We had a different problem: how do you build an intelligent booking experience when your entire dataset is "about 3,000 rides per year" and your tech team is one developer?
The constraint that shaped everything
Our site, navetteforyou.com, serves eight languages. We built it as a static site — no server, no database, no session state. Every page is pre-rendered HTML that loads in under one second from a CDN edge node.
That rules out server-side AI. No real-time model inference, no dynamic pricing algorithms, no chatbot backends. Everything has to work at build time or in the browser.
What we actually use AI for
1. Build-time content generation with human review
Our multilingual content goes through a pipeline: English source → AI translation draft → native speaker review → static build. The AI draft handles the mechanical translation; the human catches cultural nuance.
For Marrakech-specific content, this matters enormously. A direct translation of "the old city" misses that French speakers say "la médina" and German speakers say "die Altstadt von Marrakesch." The AI gets close; the human fixes the last 10%.
2. Structured data from unstructured routes
Morocco does not have standardised addressing. Our pickup points are described in natural language: "the small door next to Café des Épices, Rahba Kedima." We use an AI pipeline at build time to extract structured location data from these descriptions and generate Schema.org markup that search engines can parse.
The result: Google understands our service area without us manually geocoding every riad entrance in the medina.
3. Static pricing intelligence
We pre-compute every route price at build time. The AI component is not in the pricing itself — that is a business decision — but in the validation layer. A script checks every rendered page for pricing consistency: no route should cost less than our floor, no airport transfer should show a price that undercuts our published rates.
This catches the kind of error that a human reviewer misses when scanning 641 pages across eight languages. The check runs in CI — if it fails, the deploy is blocked.
4. WhatsApp-first booking flow
Most of our customers book via WhatsApp, not a web form. The site generates deep links with pre-filled messages that include the route, date, and passenger count. The "AI" here is minimal — template matching and URL encoding — but it bridges the gap between a static website and a conversational booking channel.
The result is a booking funnel that feels intelligent to the customer (the message arrives pre-formatted with their trip details) while running on zero infrastructure.
What we deliberately do not use AI for
- Dynamic pricing: Our prices are set by the business, not by demand curves. We do not surge-price airport transfers when a flight lands at midnight.
- Chatbots: A WhatsApp message to our driver is faster and more accurate than any chatbot we could build.
- Review analysis: With a few hundred reviews, we can read every one. Sentiment analysis solves a scale problem we do not have.
The takeaway for small businesses
The AI discourse is dominated by companies with massive datasets and dedicated ML teams. But the most useful AI applications for a five-vehicle transport company are mundane:
- Translation drafts that cut localisation time by 80%
- Validation scripts that catch errors across hundreds of pages
- Structured data extraction from messy real-world inputs
- Build-time processing that eliminates runtime infrastructure
None of this requires GPUs, training data, or a data science hire. It requires knowing which problems are genuinely helped by AI and which are better solved by a human answering a WhatsApp message.
Marrakech Transport operates private airport transfers and day trips across Morocco. Our site is built with Astro and deployed on Vercel. Book at navetteforyou.com or WhatsApp +212 701 664 704.
Top comments (0)