AI That Reads Pricing Rules Just Cut a Tourism Team From 20 People to 3

Pricing a tour isn't simple. Every order might involve different attractions, hotels, discounts, and special clauses — and the rules change often. A Chinese tourism company used to have a team of 15 to 20 people manually working through each order, spending about 10 minutes per order. That's slow and expensive.

So researchers built a hybrid AI system. Instead of letting AI make the final price (which could be risky or wrong), they split the job. The AI part reads the order and the policy documents, then picks which rules apply. A separate, traditional computer program does all the actual number-crunching. That way, you get the flexibility of AI with the reliability of a calculator.

The policies themselves are turned into something like decision trees — clear, branching structures that humans can inspect. That means new rules can be added without rewriting software, and every decision leaves an audit trail. If someone asks, "Why was this price $437?", there's a clear answer.

In a real deployment across 7 scenic sites and 12 business categories, the system handled 3,960 orders over six months. The order management team dropped to 3 people, and each order now takes under 2 minutes. The researchers also fine-tune the AI periodically using past orders, so it gets better at matching the right rules over time.