When we started Exon Labs, we made a deliberate choice not to build another horizontal AI tool. The market has plenty of those. We wanted to apply AI where the gap between user expectations and current product quality was widest — and where the data available to train on was richest.

Travel met both criteria. Booking a trip today requires navigating three to five separate products: a flight aggregator, a hotel site, a reviews platform, a mapping tool, and often a social network to validate the destination. None of these talk to each other. None of them know anything meaningful about you.

The data landscape is also unusually rich. Every booking is a structured signal: what was chosen, what was rejected, what price was accepted, what destination recurred. This is exactly the kind of dense, transactional data that AI models use effectively. The challenge has been that existing platforms have treated this data as a retention moat rather than a product asset.

We also looked at the frequency of the problem. International leisure travel happens a few times a year for most people, but domestic and business travel for frequent travellers happens every week. High-frequency decisions with high stakes are where intelligent tooling delivers the clearest return.

The honest answer is also that we find the problem interesting. Travel decisions involve real trade-offs, real constraints, and genuinely personal preferences. Building AI that reasons well about that is a hard problem. Hard problems produce better products.