Peter Jarvis, Jonathan Moore, Jussi Stader, Ann Macintosh, Andrew Casson-du Mont, and Paul Chung
In this paper we describe how we are exploiting AI technologies to infuse workflow systems with adaptive capabilities. This work is part of an ongoing applied research programme between AIAI and a number of industrial and academic partners. We begin by presenting the requirements of adaptive workflow within a taxonomy consisting of the layers 0f domain, process, agents, organisation, and infrastructure. We then show how each level can be substantially addressed with AI technologies. Specifically, infrastructure adaptation is addressed with multi-agent toolkits, agent adaptation through knowledgebased capability matching, organisational adaptation through authority based capability matching, process adaptation through AI planning and execution architectures, and domain adaptation through rationale capture. We conclude by identifying important challenges for further work as being the improvement of rationale capture and the support for the evolution of the process models that underlie executing processes.