New AI Predicts How Your Order or Claim Will End

When you file an insurance claim, track a package, or open a customer service ticket, you're watching a process unfold. Companies want to know early how it will end: Will this claim be approved? Will the delivery be late? Will this repair take three days or three weeks? A team of French researchers has published a new method for answering that question, and it beats older approaches in their tests.

The clever part is how it looks at the situation. Most existing systems read a process like a sentence — step one, step two, step three — and try to guess the next words. That falls apart when things get messy, because two customers can follow very different paths and still end up in the same place. This new approach instead draws the whole situation as a map, with events as points and timings as connections. The AI then searches for past situations that look similar, like a librarian finding the closest matching case file instead of guessing word by word.

In plain terms, it's less about "what happens next" and more about "what usually happens to people in this exact spot." That matters for accuracy, because it can weigh both what happened and how long each step took. The researchers tested it on two real business datasets and found it produced the most accurate endings overall, both in the order of events and in the timing.

The catch: this is a research paper, not a product you can use. It was tested on only two datasets from specific industries, so it's unclear how well it would hold up across wildly different businesses. And any system that predicts outcomes from your data raises privacy questions about who sees it and what decisions get made from it.