Agile system development life cycle (SDLC) focuses on typical functional and non-functional system requirements for developing traditional software systems. However, Artificial Intelligent (AI) systems are different in nature and have distinct attributes such as (1) autonomy, (2) adaptiveness, (3) content generation, (4) decision-making, (5) predictability and (6) recommendation. Agile SDLC needs to be enhanced to support the AI system development and ongoing post-deployment adaptation. The challenge is: how can agile SDLC be enhanced to support AI systems? The scope of this paper is limited to AI system enabled decision automation. Thus, this paper proposes the use of decision science to enhance the agile SDLC to support the AI system development. Decision science is the study of decision-making, which seems useful to identify, analyse and describe decisions and their architecture subject to automation via AI systems. Specifically, this paper discusses the decision architecture in detail within the overall context of agile SDLC for AI systems. The application of the proposed approach is demonstrated with the help of an example scenario of insurance claim processing. This initial work indicated the usability of a decision science to enhancing the agile SDLC for designing and implementing the AI systems for decision-automation. This work provides an initial foundation for further work in this new area of decision architecture and agile SDLC for AI systems.
翻译:暂无翻译