Spent 6 hours arguing with a chatbot about my insurance claim

I had a fender bender last Tuesday and tried to file a claim through my insurance company's AI chatbot. It kept asking me for my policy number in three different formats and then told me my coverage didn't exist for 20 minutes. Turns out the bot was trained on data from a different state's policies entirely. How can we trust these systems with stuff like healthcare diagnoses when they can't even handle a basic car accident?

2 comments

2 Comments

derek_schmidt619d ago

Honestly, the part about the bot being trained on a different state's data hit home for me. I once spent two hours talking to a customer service bot about a refund for a pair of boots I ordered, and it kept asking if I needed help with my "home kitchen appliance." No clue where that came from, but it just kept looping and looping until I almost threw my phone across the room. Ngl, if they can't even get the state right on your insurance, I don't trust them with anything serious like medical stuff either.

mark_thomas19d ago

The bot in that boot story sounds more like a simple glitch than a training error. Insurance chatbots are supposed to be built around specific state regulations and policy types. If yours was pulling from another state's data, that's a pretty serious programming mistake. The boot bot just sounds like it had a bad word association or a code bug. Those are two different problems really. One is a training data failure, the other is more like the software just broke in a weird way. Still, neither one gives me much confidence in these systems handling anything beyond simple yes or no questions.