Here's one argument against it, for devil's advocate sake. Sometimes, when doing performance tuning there are multiple ways to refine a particular query that can lead to different execution plans. In Query By Example, how would these VERY different statements be represented? :

-- LEFT JOIN syntax will do a 

-- MERGE JOIN if indexed properly

SELECT Column1 FROM Table1

LEFT JOIN Table 2 ON Table1.PK = Table2.FK

WHERE Table2.FK IS NULL

-- SUBSELECT syntax will do

-- a NESTED LOOP JOIN if indexed properly

SELECT Column1 FROM Table1

WHERE Table1.PK IN (SELECT Table2.FK FROM Table2)

-- EXISTS syntax will also do

-- a NESTED LOOP JOIN or better JOIN if 

-- inedexed properly

SELECT Column1 FROM Table1

WHERE EXISTS (SELECT * FROM Table2 WHERE Table2.FK = Table1.PK)

The point is, Query By Example won't help identify performance gains from slight syntax variations that Query Analyzer coding will.