For any of you who are adopting GenAI, what factors are influencing the choice of initial use cases?

Risk is one I encounter often. Customers pilot GenAi on internal use case be it conversational search or agent assist type use cases 

It is early in the game and there is tons of low hanging fruit. Quick wins can be found by looking at high volume tasks or workflows that are well suited to models like Claude2 (AWS Bedrock) - for example, we mark up IT infrastructure events instantly with steps to validate, troubleshoot, remediate and even insights as to what the events mean and what their likely business impact is. You can play with this at http://cascadeo.com by signing up for Cascadeo AI (which is free of charge) to get a sense for just how powerful and effective this is. Other easy quick win use cases are for common contact center operations like responding to routine questions or requests from end customers. This is vastly better than the useless chatbots of a year ago and in most cases is indistinguishable from a human agent.