The performance of an agent depends highly on its ability to adapt its methodologies to solve different problems, or to solve the same problem in different situations. Such adaptations may (or may not) change the external behavior of the agent. This paper introduces two adaptive methodologies that utiliTed by intelligent agents for improving their performance. In the first methodology, the agent adapts existing knowledge-base to fit different decision making situations. The agent restructure the knowledge-base when the user requests information or a decision that can not be directly determined from the knowledge-base. The agent uses a set of cost functions and other parameters to describe different decision making situations. A simulated air-travel database is used to illustrate the adaptive methodology. The second agent generates optimized plans for object recognition. The goal of the agent is to recognize visual objects in order to conduct the appropriate actions. The agent should have many tools or classifiers, each utilizes different parameters or recognizes different characteristics of the objects. Initially, the agent generates a recognition plan. Then, it iteratively adjusts the plan until it satisfies a set of criteria that maximize the recognition rate. A set of real-world hand gesture images is used to illustrate the adaptive methodology.