In general, VoIP and interactive video should be QoS treated to minimize latency and/or drops. (On Cisco devices, it's this traffic that is generally placed in LLQ or PQ.)
VoIP control, video streaming, management (BTW, why kind of management traffic?), often don't need the same latency guarantees, as above, but they too often need guarantees to minimize drops.
Most traffic (including that in the preceding paragraph's) works well with best effort treatment assuming there's flow fairness. Without it, you end up needing many classes with bandwidth guarantees for each class.
Oh, and although most think of QoS for providing "better" treatment for some traffic, often improvement can be made also by providing "worse" treatment for some traffic, such as having a "scavenger" class. Placing suitable traffic in such a class can remove "pressure" against other BE traffic.
