When developing an agent system, it is vital to consider the issues that have been mentioned below. They are guidelines that previous unsuccessful multi-agent systems present to future developers.

• Overselling agent solutions or failing to understand where agents may be useful:
  By this we mean that a clear planning strategy needs to be devised before an implementation of a multi-agent system is considered. In addition, it is best to weigh the pros and cons of implementing the proposed application as a multi-agent system in comparison to the primitive software solution approach.


• Vague understanding as to why agents may be required:
  This requires a careful job description regarding the different tasks that an agent is supposed to be able to deal with. For instance, a businessman may require an agent system in order to maximize efficiency. However, the businessman may not realize that the agent's efficiency can be maximized by outlining multiple tasks such as forecasting as well as further product developement.


• Design does not involve concurrency:
  By this we mean that the scale at which concurrency takes place should be maximized in order to gain the benefits of implementing a multi-agent system. Previous systems have been unsucessful due to their low levels of concurrency.


• Agents interact too freely or in a disorganised way:
  It is best to keep the number of agents down to a minimum level. This is due to the fact that the more agents there are, the more interaction is required. In addition to this, multi-agent systems require a systematic method of engineering in order to ensure that interaction takes place in the correct manner, without chaos.


• Building an agent architecture:
  When designing multi-agent systems, it is necessary to decide on the architecture of the agent system. However, it is advised that a pre-defined architecture template is adopted and the agent system is built on that. Many systems have failed due to the idea of creating a customised agent architecture.

These are the major pitfall areas that a Multi-Agent programmer can experience. However, by considering the points mentioned above, the success rate of the agent system that is to be implemented may be raised.