arrive at the ideal solution. In scenarios with high volumes but simpler tasks, configurations 5 and 6 might work well. Conclusion GenAI, primarily in the form of large language models, has already begun transforming customer support. From travel and tourism to healthcare and education to financial services, chatbots based on LLM’s have the potential to make customer support a higher-quality experience for the customer while also improving the efficiency for the organization providing it. In our discussions and experience in several industries, we were surprised to see that the biggest challenge in the transition of GenAI tends not to be a legal or technical challenge but rather for management to develop a vision of how GenAI could be deployed. Before starting the technical development of a chatbot, executives need to ask themselves what type of bot they would like to get. The goal of this paper is to help executives develop such a vision by reimagining customer support. We have presented five dimensions of chatbot designs and six configurations specifying workflows how humans and GenAI collaborate to provide customer support. Our five dimensions of bot design and our six configurations together create a menu of design options. This menu can be used for ideating the needed vision. Should an organization work on a chatbot that has broad knowledge, recognizes a customer over a longer relationship, and is able to learn from past interactions and then deploy it by having a human operator interacting with the customer while getting real-time LLM-powered oversight? Or should the organization pursue a chatbot that is focused on particular support problems, has no memory of past interactions, but is able to autonomously interact with customers with only episodic quality auditing? In our view, there is not one best design or workflow configuration. Instead, it is management’s role to systematically explore the potential designs and configurations and use the business objectives described above to find the most promising customer support vision for their own organization. Acknowledgements We thank Vibhanshu Abhishek, Martin Bittner, Lilach Mollick and Hummy Song for their helpful comments.