[pizza234]

I think indexing is ultimately a trade-off design; the end effect of it is a decision which belongs to a human - what do we make faster at the expense of what?

Probably the best thing a "machine" can do is to give tools to easily perform a reasonably accurate and extensive analysis, but I think any major RDBMS has "good enough" tools to do so.

Interestingly, I think it's important to think who is the target of such tools - the casual developer? the novice dba? the expert dba who works on large datasets? I think even theoretically advanced tools would help only the first case; the other two need to know very well what they're working with.