Part of the day job at the moment is talking to both colleagues and clients or potential clients about using AI in their work. That might mean looking at how we could use AI to help with software development, how it can be used when writing presentations or even how it can be used in business process automation.
Something I’ve noticed when trying to get people enthused about the new opportunities AI represents is that the popular tools that are available are often seen as toys rather than tools.
People try out something like Chat GPT or one of the image generation tools. They have a lot of fun playing with it, but don’t go deeper into what it can do and then lose interest, deciding that while fun it’s not useful for serious business applications.
Getting the best out of LLMs requires a lot of trial and error and a lot of prompt tweaking and iterations. The commercially available tools are designed to have a low barrier to entry, but most have a lot more depth that isn’t immediately obvious unless you spend a lot of time researching them. I think that design approach has led to explosive growth in the number of people trying them out, but I think it’s causing some pushback against using them for more serious applications.
I don’t think the solution is to change the end user products like Chat GPT, but I think there is an opening here for applications designed for business use that surface more of the depth and complexity of the underlying models.