It is well known that the use of multiple antennas can enhance the capacity of wireless links delivering messages to geographically dispersed receivers. Moreover, in applications where the channel state information is available at the transmitter, efficient precoding strategies (based on information embedding concepts) exist for multiplexing the different messages to be delivered to different users. In particular, this precoding performs an interference pre-cancellation operation that ensures the different target receivers can ignore interference effects when decoding their messages. However, in typical networks, the collection of messages to be transmitted can be quite large, and a scheduler selects a subset to be multiplexed in any particular transmission interval. When the scheduler and multiplexor operate independently, the scheduler complexity is very low, but the multiplexor can require substantial complexity. We show that by having the scheduler make efficient use of a latency window and take into account the implications of its decisions for multiplexing, the overall system can be much more efficient. In particular, the complexity of multiplexing is significantly reduced while requiring only a small increase in the complexity of scheduling. In addition to highlighting the role of delay as a resource, this work is another compelling illustration of how a modest amount of lower-layer awareness at the networking layers can substantially improve overall resource efficiency from a network architecture perspective.