Over the years, I've often chuckled under my breath when a client would tell me: "The system is giving me different answers to the same query. I think it's making thing up!" In almost every case, what the client perceived to be "the same query" was in fact two different queries. The underlying logic was different. Sure, they both said "Member count" but one was counting only active records while the other was counting all records (or something to that effect).

But AI queries are different, because as we all know, AI WILL make things up (hallucinations!).

And so, at the moment, one of the greatest deterrents to AI adoption within the AMS is the potential for it to make up data! As someone put it to me recently, "The first time it lies to me is the last time I'll use it."

There are still plenty of other ways AI is and will be useful within the AMS. But solving the hallucination challenge will be critically important. I don't think we're there yet.