I was on a call this morning where a team had decided to break it to their leadership that genai was not the answer they thought it was going to be. I bet there's going to be a lot of that this year.
.net all over again.
Microsoft is so incredibly bad at marketing. And really not great at products too. People don't pick them because they're the best at what they do, but because they're what everyone else uses and they're acceptable. The whole reason for Teams being adopted so heavily is because it comes with office and it's not bad enough to justify paying for something better. But when having used other products that were actually leading in their class, it's sometimes sad being stuck with it.
In this light I do understand their rush to capture the AI market, being first is just how they work.
If you can't guarantee the output of the llm how can you ever expect this to always work the same? How can you even test it? It may test out ok 99 times but on test #100 it fails. Any hallucination breaks it or, even worse, produces a confident sounding wrong output. Is that the reliability problem you're talking about?