Paper Number
2082
Paper Type
Complete Research Paper
Abstract
While chatbots can be implemented with very little effort, scaling and maintaining chatbots remains a challenge. This is crucial in knowledge-intensive customer service like IT support, where domain knowledge must stay current with the evolving IT landscape. Following design science research, we derive design principles for a generative AI (GPT4) enabled textual training data creation and curation system (T²C²) as part of a new class of systems – bot delegation systems. For the design of T²C², chatbot and domain expert viewpoints are integrated. We evaluate two instances of T²C², each with distinct degrees of human-ai delegation where employees act both as creators and curators of training data. The paper’s theoretical contribution is two-fold: (1) we present a novel kernel theory that represents the material characteristics of bot delegation systems by contextualizing the IS delegation framework to the self-determination theory; (2) the design and evaluation of T²C² as the built-and-evaluated artifact.
Recommended Citation
Reinhard, Philipp; Li, Mahei Manhai; Peters, Christoph; and Leimeister, Jan Marco, "Let Employees Train Their Own Chatbots: Design of Generative AI-Enabled Delegation Systems" (2024). ECIS 2024 Proceedings. 5.
https://aisel.aisnet.org/ecis2024/track23_designresearch/track23_designresearch/5
Let Employees Train Their Own Chatbots: Design of Generative AI-Enabled Delegation Systems
While chatbots can be implemented with very little effort, scaling and maintaining chatbots remains a challenge. This is crucial in knowledge-intensive customer service like IT support, where domain knowledge must stay current with the evolving IT landscape. Following design science research, we derive design principles for a generative AI (GPT4) enabled textual training data creation and curation system (T²C²) as part of a new class of systems – bot delegation systems. For the design of T²C², chatbot and domain expert viewpoints are integrated. We evaluate two instances of T²C², each with distinct degrees of human-ai delegation where employees act both as creators and curators of training data. The paper’s theoretical contribution is two-fold: (1) we present a novel kernel theory that represents the material characteristics of bot delegation systems by contextualizing the IS delegation framework to the self-determination theory; (2) the design and evaluation of T²C² as the built-and-evaluated artifact.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.