In such Expert Systems, the expertise of the Expert System is derived from two components of the Expert System: the knowledge base and the inference engine.

The knowledge base embodies all of the Expert System's knowledge in the form of rules. A typical rule-based expert System may contain a few thousand rules. These rules provide the strategies, conclusions, directives, etc. that allow the Expert System to do its problem solving in logical steps.

All rules consist of a condition and an action to perform if the condition is true. An example of a rule may be "If there is a smoke, light and heat, then there is a fire". The condition portion of the rule usually corresponds to a fact, or a set of facts, that can be found in a database (a set of facts that describe the situation to the expert system, such that the expert system can use these facts to aid it in its problem solving). The action portion of the rule can specify new fact(s) to add to t he database (such as in the above-mentioned example), cause other rule(s) to be tested, or can perform actions that directly affect the real world, such as calling the fire department. Rules may also include a certainty factor that represent the confidence level that a fact is true or valid.

The inference engine contains the problem solving methods of the Expert System. These methods match the facts contained in the database with the condition(s) of the rules stored in the knowledge base so that conclusions about the situation can be made. The successive matching of rules (and the execution of the specifed actions) based on known facts and new facts added to the database during the problem solving process produces inference chains:

[diagram of inference chain]

The inference chains generated by the expert system during its problem solving shows us how the expert system came to its conclusions, and can be displayed to the user of the expert system to justify the expert system's decisions.