As a general rule I suspect that for every automation induced error someone can find an equivalent pilot error to balance it and vice versa. It will never be possible to prevent all fatal crashes but I think most would agree that the goal is to reduce such crashes to as low as possible bearing in mind the economics and practicality of such interventions.
What automation offers is consistency. Over the years there have been many great saves by pilots but equally there have also been many failures by pilots, be that through fatigue, inattention, misunderstandings, unprofessional behaviour or even downright recklessness. I have no stats but I strongly suspect the failures vastly outweigh the saves. OTOH I am sure there are many automation failures that might have resulted in fatal crashes if it were not for the human pilots intervening. This means that any aircraft with a reduced number of pilots will need rigorous testing but also that rigorous controls and testing are needed over software updates.
We can be certain that at some point a single pilot or autonomous airliner will crash through programmer error. The acid test will be if the average number of fatal crashes per year is considerably reduced through further automation.