Article Preview
TopBackground
In this section, we describe related studies. In this study, we considered projects that have developed AI service systems. Therefore, our survey presents related research on AI projects, AI system architectures, and AI system project models.
There are many challenges in the requirement, design, implementation, and test phases in AI system development projects (Domingos, 2012; Sculley et al., 2014; Kumeno, 2019; Lwakatare et al., 2019). One of the challenges encountered in PoC projects for developing AI service systems is the common understanding of the developed system and the development activities between stakeholders (Hofman et al., 2017; Washizaki et al., 2019).
To understand the developed AI service system, some architectures for representing the entire system in a practical project while applying Big Data analytics or machine-learning technologies have been proposed (Demchenko et al., 2014; Earley, 2015; Heit et al., 2016). However, these are mainly used by development teams. Although an architecture containing related business elements was proposed (Geerdink, 2013) so that business teams can understand the developed AI service system, it did not contain the project goals or main people involved in each element of the project. Therefore, it is difficult to use this architecture for discussing AI service system development projects between business and IT divisions.
A multiviewpoint evaluation method for an IT system is the business–IT alignment. In enterprise system management, business–IT alignment is introduced and defined as the relations among the business goal, business process, and IT system for tasks such as decreasing organization uncertainty and improving enterprise agility (Zhang et al., 2018). Hinkelmann et al. (2016) and Saat et al. (2010) introduced methods for constructing a business–IT alignment model by using an EA modeling approach and analyzing it continuously in a company.