We Built an AI Travel Concierge. Here's What Happened.
The story of building an AI assistant that actually knows your miles
I was standing in the United Club at O'Hare, 90 minutes before a flight to Tokyo, staring at two browser tabs. One showed my Chase UR balance (85,000 points). The other showed United award pricing for Polaris business class (80,000 miles). I had a simple question: should I transfer to United, or would my points go further with Hyatt?
I spent 20 minutes searching blogs, checking transfer ratios, comparing hotel prices. By the time I figured it out, they'd called boarding.
That moment, standing in an airport lounge, drowning in browser tabs, trying to do math that a computer should be doing for me, is exactly why we built the MileIntel Concierge.
Not Another Chatbot
Let me be direct: the world does not need another AI chatbot. What it needs, what I needed in that United Club, is an assistant that has context.
The concierge isn't a general-purpose AI with a travel skin. It has access to your actual loyalty balances. It knows current award pricing across programs. It tracks transfer partner ratios and knows when bonuses are active. It can compare redemption values in real time because it's connected to the same data pipeline that powers all of MileIntel.
When you ask "Should I transfer Chase points to United or Hyatt?", it doesn't give you a generic answer. It gives you your answer, based on your balances, current pricing, and the specific trip you're planning.
The Hard Part Wasn't the AI
Here's something that might surprise you: building the AI part was straightforward. Claude is remarkably good at understanding loyalty program questions and structuring comparison data.
The hard part was the data pipeline.
We're tracking 47 airline programs and 12 hotel programs. Award pricing changes daily, sometimes hourly during flash sales. Transfer ratios shift. Programs add and remove partners. Some programs have published award charts; others use fully dynamic pricing that requires monitoring.
Each program needs its own data pipeline:
The AI is the interface. The data infrastructure is the product.
The Travel Command Center
The concierge is one piece of a bigger picture. When you open a trip in MileIntel, you get the Travel Command Center, a live operational view of your journey.
Think of it as your personal flight ops dashboard. It shows your flight status, gate information, lounge access, destination weather, and the concierge, all in one expanding panel. You don't need five different apps open. You need one screen that knows what matters right now.
What We Got Wrong (At First)
We originally built the concierge as a simple Q&A bot. You'd ask a question, it would search our database, return an answer. Clean, simple, boring.
The problem was that people don't think in single questions. That United Club moment wasn't one question — it was a chain: What are my options? What does each cost? What's the value? What about the hotel? Should I split points across programs?
So we rebuilt the concierge to maintain conversation context and chain tool calls. It checks your balance, then pricing, then calculates per-point value, then factors in transfer bonuses, all in one exchange. It takes about 4 seconds to do what took me 20 minutes with browser tabs.
Try It
The concierge is live for all MileIntel accounts — free tier included. Ask it anything about your miles. The question I'd start with: "What's the best use of my points right now?"
It has the context most chatbots don't. That's the whole point.
The best time to use your miles is when you know exactly what they're worth. The concierge makes that instant instead of a research project.
This is a From the Cockpit post, where we share what we're building, why, and what we learned along the way. Create a free account to try the concierge yourself.
Don't Miss a Departure
Track your miles, catch devaluations before the blogs do, and find the best use of every point you have.
Start Tracking Free