Every day, information is transacted in the Internet across different enterprises in business-tobusiness (B2B) collaborations. In a bid to manage the growing volume and complexity of supply chain information, many companies are turning to business process management systems (BPMS) and systems supporting B2B collaboration standards such as Electronic Data Interchange (EDI), RosettaNet and ebXML (Electronic Business using eXtensible Markup Language). According to a 2007 Gartner report, the BPMS market reached nearly US$1.7 billion in total software revenue at the end of 2006. This is testimony to a rapid growth of interest in business process management (BPM) over the last decade. Despite this growth, many current techniques are still static in nature, requiring substantial modifications/ (re)coding to re-establish the link between collaborating companies. Such modifications are expensive and impractical in a dynamic, global business environment characterized by frequent, rapid changes. At the same time, many prominent business process (BP) modelling techniques are rooted in computer process theories like state machines, Pi Calculus and Petri Nets, which are computationally efficient but are void of context and semantics so necessary in very complex, multi-party, high level business communications. Most importantly, these BP modelling techniques cannot dynamically synthesise business processes based on a user's high-level business goals and constraints. The goal of this thesis is to discover, implement, test and verify a technique that enables agile B2B collaborations through the dynamic formulation of cross-enterprise business processes based on high-level user business goals and criteria. An extensive literature review into current BP research, industry best practices, BP modeling standards and B2B industry standards, revealed that the realization of this goal is contingent upon solving four sub-problems: (1) the problem of decomposing high level business goals and constraints (e.g. item price, lead time) on-the-fly into corresponding low-level tasks of adequate granularity, (2) the task of dynamically chaining these decomposed low-level tasks with the appropriate control and information flows, (3) the problem of representing the semantic characteristics and relationships between business processes components (e.g. tasks, sub-tasks, actors, methods of decomposition, control flow, etc.) and (4) the chained low-level tasks have to be implemented openly and directly on the WWW. These four sub-problems were addressed by a hybrid methodology which entails the dynamic decomposition and sequencing of tasks facilitated by Hierarchical Task Network (HTN) planning concepts, and the adoption of ontologies. HTN Planning employs so-called methods to establish the various permutations by which high level business processes can be © Ryan K. L. Ko, 2011 vi decomposed into primitive tasks which are chained sequentially in a single level from left-toright. Current business processes in industry do not use these methods. However, t...