> This worked sometimes (I’d estimate >98% of the time), but failed enough that we had to dig deeper. > While we were investigating, we noticed that another fi

> This worked sometimes (I’d estimate >98% of the time), but failed enough that we had to dig deeper. > While we were investigating, we noticed that another field, name, was consistently returning the full name of the state…the correct state – even though we hadn’t explicitly asked it to do that. > So we switched to a simple string search on the name to find the state, and it’s been working beautifully ever since. So, using ChatGPT helped uncover the correct schema, right?