We investigate the suitabilility of the KGP (Knowledge, Goals, Plan) model of agency for autonomous decision making in dynamically changing environments. In particular, we illustrate how this model supports the decision making process of an agent at different levels, while the agents generates goals, plans for these goals, and selects actions to achieve the goals that it has planned for. We also exemplify the approach by illustrating how the model and a prototype implementation in the PROSOCS platform can be adopted to support e-negotiation, using a particular kind of internet auctions as a case study.\r\n