Having worked with AI on-and-off for decades, I wouldn't expect it to do interpretation and diagnosis itself (it's bizarrely wrong too often, and we can't even trace back and figure out why); instead, it's useful for turning up extra things for humans to evaluate by spotting patterns in huge bodies of data (more than a human could look at).

I tell people to treat AI like an over-enthusiastic intern who's always sticking a head into your office saying "look what I found!" "look at this!".

Most of the time you'll just smile politely and try not to roll your eyes, but every once in a while it will turn up something genuinely useful.