I'd agree with others that the Big Five is the most empirically validated, but it's also the most flat and flavorless. It entirely lacks the the combinatorial factor that makes the MBTI interesting. Still, using google scholar with the Big Five yields a treasure trove of studies, which is great if one is looking for empirical evidence. One can relate those studies back to the MBTI (which adds in a little flavor), but looking at correlations of correlations makes things extremely tenuous.
At this point, I'd say I don't really buy into MBTI type dynamics (although there's a chance Nardi-esque research could redeem aspects of it), and think "the functions" are more an effect of preference combination (so Fi = F + P, etc). I kind of agree with Reynierse, that MBTI preferences are continuous, and one's strongest preferences (in pairs or triplets) have the greatest effect. MBTI's test/retest statistics are right up there with Big Five when preferences are viewed as a continuous scale, rather than as a dichotomy.
The enneagram I find personally very useful, but mostly for personal growth and getting insight into the ways we sabotage connection and relationships. I like that it provides a not-overly-pathologizing perspective on habitual defense mechanisms that encompasses people at various levels of mental health. I don't find the particular layout of the Enneagram and the lines of connection particularly inevitable or convincing. It seems likely to me that people shift defensive strategies when exhausted or frustrated, but which one they shift too seems just as likely to be the result of personal or environmental factors. I think it's the least scientific, hardest to create instruments for, and most difficult to study empirically of the four (Big Five, MBTI, The Ennegram, and Socionics).
Socionics I have the least use for. Not only does the problem domain overlap with the MBTI and Big Five significantly, but most of the in-depth books and studies are unavailable in English (I assume they exist). That makes it hard to evaluate studies and get a deeper view into the system. The visual identification also seems ridiculous, although I could see particular expressions or movement patterns being useful. I haven't found any socionics type a better fit for me than types in the other systems offer, so at this point it seems a personal lose/lose/lose.
As far as domains of utility: I find the MBTI useful at work in particular (and sometimes in social situations), where it helps give me a framework for understanding people's mental blind spots, and for appreciating strengths, talents and perspectives that are foreign to me. Conversely, the Enneagram has been useful for me in therapy and close relationships, and I appreciate the unpleasant insight that it provides. I don't find it particularly useful in work contexts (unless folks are particularly dysfunctional), and have found attempts to apply the enneagram to work contexts to be fairly unconvincing. Big Five is useful to me for tracking down empirical research.