Raw LLM Responses

I've done some contract work doing LLM integrations for a non-tech company. As the built in integrations get better and better, more and more 'non-technical' users will have a good use case for LLM's as well as easy access to it. In my opinion, we're in the early phases of accessibility right now for non-programmers. Coding integrations were tackled pretty quickly and imo the integrations are more straightforward, it's baked into your IDE and can access your files. In non-tech corporate, workers have all kinds of information and tools it needs to access and edit- excel files, emails, sharepoint files, powerBI, reports, whatever. Right now those workers are largely in the copy / paste phase. I unfortunately believe that AI will become baked into most of the products that workers use and the workflow will be more 'native.' They'll go from the programmer equivalent of copy pasting into chatgpt circa 2022 to having opus running in vscode. They're rolling out integration features pretty fast. I think it's only a matter of time before some kind of new excel / spreadsheet program launches (or they upgrade excel / sheets) to have AI baked into it (and I mean one that works as well as opus baked into vs code). Claude and OpenAI are already able to hook into a lot of microsoft products (outlook, sharepoint, etc) and it fairly intelligently summarizes and parses information. I've definitely seen people at the org I contract for have their labor reduced. So to answer your question, and this is just my opinion, enterprise customers will pay for it. And their workers will use a lot of tokens, just like SWE's, once they start making it more accessible. They'll probably use more tokens if I had to guess.