The theoretical background presented in the thesis was divided into three chapters: supply chain management, intelligent agents, and COOL. The first chapter presented some important issues of supply chain management with focus on the interrelations among flexibility, inventory management, and customer satisfaction. It is stated that integration and coordination are important tools in improving supply chain management. The second chapter introduces a coordination expert, the agent. An agent is an autonomous piece of software with knowledge of its environment, the ability to react to changes therein, and the ability to communicate with other agents. It is argued that the supply chain is a well suited application domain for multi-agent systems.
The COOrdination Language, COOL, is an agent language developed at EIL. COOL provides constructs for defining agents and conversation classes. The conversation classes are rule based coordination plans allowing the agents to coordinate their behavior in complex ways. The social ability of the COOL agents is assured by message passing, with the use of KQML.
The practical work described how COOL was used to design and implement two supply chain models. The aim of the implementation of the HP's Simple Model  was for the author to become familiarized with COOL, and the notion of running simulations on a multi-agent system. The implementation was verified by comparing simulation results with those obtained at HP-labs.
The second and major part of the practical work involved the modeling of the fictitious Perfect Minicomputer Corp (PMC). The agent approach proved to be supportive of the designing process. Through the encapsulation of naturally occurring entities into agents, the approach facilitated the conceptualization of the system. The autonomous nature of the agents, and their very clearly defined interfaces, makes the system flexible. The relevance of the model as a supply chain model was verified through simulations. The results indicated that the behavior of the model are as would be expected from a real life supply chain.
Simulations on the PMC Model also emphasized the importance of information sharing in the supply management process. When measuring different coordination strategies' impact on RPI levels, the coordinated production planning was shown to deal with unforeseen events better than a less coordinated solution. A coordination protocol where a workstation agent shares capacity information with the planning agent in case of machine failure, was shown to enhance planning's ability to control RPI levels. The simulations gave us the opportunity to measure the impact of absence of information sharing in quantitative terms.
The objective of the thesis (as interpreted in section ) was met through the construction of the PMC Model. The PMC Model is a design of information and work flows for supply chain coordination.
It also gives the Integrated Supply Chain Management Project a new supply chain demonstrator. Both the functionalities and typologies of the agents are extended from the prior to the new demonstrator. The new agents provide for information and work flow requirements that are more consistent with those of a real life supply chain. The second main issue of the thesis objective is thereby addressed.
The first main issue is addressed through running simulations on the PMC Model. Simulations show that an increasing degree of information sharing and coordinated production planning enhance the reactivity of the supply chain model. Information sharing gives the local agents the necessary knowledge of the overall state of the system, which allows them to make decisions that are beneficial not only locally, but for the supply chain as a whole. Looking back to section we see that this corresponds to the definition of enterprise integration.
The PMC Model is still a fairly simple supply chain model. An obvious suggestion for further work is to extend the model. The model may be extended both in width and in depth. By extension in width we mean adding more supply chain attributes to the model. These may be: more plants, supplier selection and relations, currency, cost calculations, more products, more realistic distribution, and so on. The PMC Model has three levels of manufacturing, with only one plant on each level. An interesting topic for further research would be concerned with how the multi-agent system can handle the necessary coordination and information sharing for handling several plants on each level, for example by letting more of the parts be internally supplied. In depth extensions will add to the agents' problem solving tools, e.g. including real scheduling tools. A more extensive model will allow further research within the fields of information sharing and coordination, and will bring the approach a step further towards introduction to real life supply chains.
Another direction is to improve the model as a demonstrator for the ISCM project. This may be done by adding graphical user interfaces to the agent environment. Work is in progress to build interfaces allowing the use of World Wide Web browsers to browse and query agents.
To enhance the usability of COOL for modeling purposes the language may be supplemented by a library of agent types and coordination protocols. Jayashankar et al.  are, as mentioned earlier, conducting research within this field. An agent library would force an investigation of what knowledge is necessary for different agent types, facilitating the process of modeling.