In this paper, I deal with the problem of the supposed reference-shifting of the first-person singular pronoun “I”. Traditionally, indexicals shift their reference from context to context but in accordance with the rules of language. The distinction between pure indexicals and true demonstratives lies in the fact of their automatic reference. However, the so-called thesis of automatic reference of the indexical “I” has been questioned by philosophers and linguists who presented counterexamples (which I refer to as “I-cases”) where reference-shifting does seem to occur: “I am parked out back” referring to the speaker's car or “I am on a purple square / Baltic Avenue” referring to the speaker's game piece. I argue that the interpretation of I-cases can be explained without denying the thesis of automatic reference of “I”. First, I consider Nunberg's view on I-cases, which does not give up the idea of automatic reference of “I”. According to him, I-cases are examples of predicate transfer where the speaker inherits the secondary property by virtue of having ownership relations with an object. I put forward two main objections to his account. The first one concerns the semantic indeterminacy of relationships between the speaker and the object in question. The second one is about the criteria for what predicates can and cannot be transferred. Both objections are based on Nunberg's commitment to the pragmatic explanation of I-cases. Second, I review Romero and Soria's account where I-cases are understood as examples of linguistic ellipsis (“I” is a part of the reduced NP) and do not endanger the thesis of automatic reference of “I”. My criticism against their view is that it is open to the objection of semantic indeterminacy too and invokes pragmatic explanation, which results in extra demands on the hearer who needs to know more than just the semantic value of the expressions in order to interpret the utterance. Finally, I develop my own view, according to which neither predicate transfer nor the NP reduction occurs in I-cases. I propose to consider them as the cases of the VP reduction where the property the speaker predicates to herself is a property of having done a particular action in the past. Such a predicate is perfectly appropriate to apply to a human being, who has the ability to perform actions. My proposal is free from the objection of semantic indeterminacy in virtue of the fact that the sense of agency is implied in the semantics of “I”.