Answer:
Our ability to talk to ourselves and think in words is a major part of the human experience of consciousness. From an early age, individuals are encouraged by society to introspect carefully, but also to communicate the results of that introspection. Simon Jones and Charles Fernyhough cite research suggesting that our ability to talk to ourselves is very similar to regular speech. This theory originates with the developmental psychologist Lev Vygotsky, who observed that children will often narrate their actions out loud before eventually replacing the habit with the adult equivalent: sub-vocal articulation. During sub-vocal articulation, no sound is made but the mouth still moves.
Jones and Fernyhough cite other evidence for this hypothesis that inner speech is essentially like any other action. They mention that schizophrenics suffering auditory verbal hallucinations (AVH) need only open their mouths in order to disrupt the voices in their heads. To try and explain more about how inner speech works, but also what goes wrong with AVH patients, Jones and Fernyhough adapt what is known as the "forward model" of motor control, which uses the idea of "efferent copies". A forward model of motor control. Notice that a prediction of the future state is made just before the movement occurs. Presumably that efference copy is used to establish agency.
In a forward model of motor control, the mind generates movement unconsciously. While information is sent to the necessary body parts, the mind basically faxes a copy of that same information to other areas of the brain. This "efferent" copy could then be used to make predictions about upcoming movements. If the actual sensations match predictions, we experience the feeling of agency. If there is a mismatch between the body and its predicted position, perhaps due to obstructions or other cognitive disruption, no feeling of agency occurs.