Different methodologies of surveillance have different costs, capabilities, and potential risks. In theory anyone stepping outside of their door could be observed at any time but in practice personal observation is very expensive and potentially obvious. Non network cameras controlled by a plethora of individual business owners might paint on net a very detailed picture of a persons comings and goings but acquiring and correlating that data likely limits it to discovering comings and goings in a limited geographical and temporal area of interest because of a major crime.
Pervasive cheap surveillance which needn't be attached to a very expensive mobile weapons platform or be limited to just cameras could be a birds eye view of everyone's lives. Good justifiable benefits are obvious. With enough surveillance crime becomes really hard to perpetrate. We can pick out all the drug dealers and make tons of arrests before people adapt. Petty stupid crime like breaking into a car means the eye of Sauron sees you and follows you back to your house. Acts of violence that don't take place inside buildings could prompt an immediate response at least as fast as the cops are capable of dispatching a unit. Acts of violence within a unit could be detected by mikes outside on street corners if we were even more surveillance minded. Perhaps we can train it to detect the sound of a person being beat. I don't like people getting hurt do you? Acts of terrorism or mass violence are even more important to prevent. With enough smarts maybe we can flag people likely to go postal or at least notice what is happening 30 seconds before the shooting starts. 30 seconds before instead of 2 minutes after might make a HUGE difference in body count.
As great as that sounds I'm sure you can think of 100 more dystopian use cases. It doesn't do us much good to treat cars as faster horses and not bothering to consider the implications.