Years ago, I sat in a dusty office in Cusco watching my top lead guide, Alejandro, close a $15,000 luxury trekking itinerary over the phone. It wasn’t the itinerary that sold the client; it was the way Alejandro described the smell of the damp earth at dawn and the specific way he promised to brew the coca tea.
That "high-touch" soul is what differentiates a $10M+ operator from a struggling agency. But for years, we faced a brutal paradox: You cannot scale Alejandro.
If Alejandro is on a mountain, he can’t answer emails. If he’s sleeping, he can’t close a high-ticket inquiry from a different time zone. Most operators try to solve this by hiring cheap VAs or using generic AI chatbots that sound like a corporate refrigerator manual. The result? Conversion rates crater because the "magic" is gone.
Today, I’m going to show you how we’ve broken that paradox. We are moving toward the "In-Person AI" Loop. This isn't about replacing your guides; it’s about capturing their lightning in a bottle and using Large Language Models (LLMs) to distribute it across every touchpoint of your business.
The Death of the Generic Chatbot (And Why Your Clients Hate It)
Luxury travelers have an incredibly high "BS meter." They can smell a generic ChatGPT response from a mile away. If your automated response says, "We offer a variety of bespoke outdoor experiences tailored to your needs," you’ve already lost the sale.
The $10M+ operators I work with have stopped using AI to write from scratch. Instead, they use AI as a vessel for proprietary knowledge.
The goal is to create a "Brand Brain"—a custom GPT or Claude instance that has been fed the literal transcripts, voice notes, and secret sauce of your veteran guides. This allows you to scale high-touch sales without losing the "founder’s soul."
Step 1: Mining the Gold (Recording the High-Touch Moments)
You cannot train a world-class AI on generic internet data. You must train it on your data. To build an In-Person AI Loop, you need to start recording your best human moments.
The Roadmap for Data Collection:
- Sales Call Transcripts: Use tools like Otter.ai or Fireflies to record your top salesperson (or yourself) closing a deal. Focus on the objections they handle and the specific anecdotes they use.
- The "Guide Briefing" Recording: Sit your best guide down and ask them: "Tell me about the hidden spot on Day 3 that isn't in the brochure." Record their answer.
- Post-Tour De-briefs: Capture the stories guests raved about during the farewell dinner.
These recordings are your "Source of Truth." Once you have 10–20 transcripts, you have the DNA of a high-conversion brand.
Step 2: Building the "Brand-Perfect" Knowledge Base
Now, we take those transcripts and feed them into a Custom GPT or a private Claude Project. This is where most operators fail—they just upload the files and hope for the best.
To make the AI sound human, you need to give it a "Persona Constraint." I use a specific prompt structure that looks like this:
"You are Alejandro, our Lead Expedition Guide with 20 years of experience in the Andes. You speak with authority but warmth. You never use corporate jargon like 'unforgettable' or 'bespoke.' Instead, you describe sensory details—the chill of the wind, the taste of the salt. Use short, punchy sentences. When a client asks a question, answer it using one of the specific anecdotes found in the uploaded transcripts."
By feeding your proprietary scripts into the system, the AI stops guessing what a tour operator sounds like and starts knowing what you sound like.



