A rule-based system handles a finite number of rules and/or has a predefined workflow for them (e.g. decision trees).
Rule-based systems are very useful in a deterministic scenario, where the situation is predictable and repeatable.
However, in the real world, things are not that simple and a rule-based system does not work in many cases.
Rational computing systems, which mimic the human brain decision making process, are designed in order to function in real world scenarios.
Like humans, Rational computing systems should function under uncertainty, ambiguity, partial observation, dynamic environment and sometime even under wrong observations (the most challenging cases).
In order to succeed in their missions, with high probability, Rational computing systems should/must:
Actively learn on-line as information changes, and as goals and requirements evolve.
Remember previous interactions, in order to increase future success rates (learning based on experience).
Resolve ambiguity, tolerate uncertainty and wrong observations.
Interact with the surrounding environment (sensing) and with their end-users so that those users can define their needs.
Assist in 'defining the problem' by asking questions or finding additional input sources if a problem statement is ambiguous or incomplete.
In the following video AiGENT Kyna resolves ambiguity, uncertainty and handles 'wrong observations' in order to meet her target with high success rate.
One can easily imagine a Rule-based system performance under the same case.