guess RNAi would be more convenient/easier in most cases...
Correct. But like John said, it does not give you 100% suppression of gene expression like a knockout does. Another benefit is that you use a promoter of your choice to regulate expression, eg you can localize the silencing effect by using a tissue specific promoter (or simply by direct injection to the region you wish to silence). What is the organism you are working/looking to work with?
This is a very versatile and powerful technology whose implications are only beginning to be understood. For instance, Dicer processed dsRNA not only acts post-transcriptionally to suppress gene expression by hybridising to the homologous mRNA region but it also acts on a DNA chromatin methylation level to silence expression pre-transcriptionally. Futhermore, miRNAs are now accepted as being another level of gene regulation where ssRNA self-hybridizes forming a hairpin structure which upon Dicer processing also supresses translation of mRNA. Such a diversity of RNAi endogenous gene supression mechanisms (originally RNAi was thought to act only as a defence against retroviruses) suggests that we have evolved a lot machinery to regulate expression and consequently presents a whole new range of intervention opportunities, from chromatin bound DNA to mRNA.