79% are using some form of OpenAI GPT. I get that that is the easiest entry point, but are companies really that cavalier about putting their data and their customers data on someone else's hardware? I get the cloud use case, but this sounds more trusting than I would be with emerging technology.
only 26% are customizing base models in any way. To me that would be table-stakes for using LLMs in any way.
I guess the main takeaway is just how early we are in this movement, and how much work there is still to do.
(edited for formatting)