Do we have knowledge bases only because the logics are not automated enough? I.e. - is it true, that the ideal situation would be that we use the set of logical formulas as the knowledge base together with the deduction services of the relevant logic as the reasoning services for this "knowledge base". But not all the logics are automated enough and that is why we are seeking approximations and representation and reasoning services for the more restricted deductive systems.
Most logical systems in use are automated enough in principle. That is, there are algorithms which find the proof of a statement, if such a proof exists. The problem is computational complexity. Such algorithms are very, very, very, very, very slow. We only know efficient algorithms for special, easy cases.
More generally, knowledge representation is not just about logical deduction. You cannot logically deduce the fact that after Monday comes Tuesday and that tigers have legs, that is something that the machine has to be told, and it has to then store such knowledge. How do we store knowledge? That is a question of expressivity of the symbolic representation (the thing you call "logic"), and it is an issue that is largely separate logical deduction.
This is really better suited for philosophy stack exchange, but there are some questions that don't have deductive answers. When there is no deductively derivable fact of the matter as to a question, then logic and axioms can't help you. In these situations we often use inductive reasoning, tradition, or intuition. Examples include "where should I look first for my lost sock," "what axioms of mathematics should we use," and "is Bob likely to be lying to me when he says he didn't eat my cookie." Also many legal and moral quandaries may fall into this category, depending on the moral or legal system.