The link I posted above is a great option that addresses the main issue: that we can never programmatically account for all cases (Smith-Jones, von Haussen, John Smith M.D.), at least not in an elegant manner. Tony introduces the concept of an exception / break character to deal with these cases. Anyways, building on Cervo's idea (upper all lower chars preceded by space), the replace statements could be wrapped up in a single table based replace instead. Really, any low/up character combination could be inserted into @alpha and the statement would not change:
declare @str nvarchar(8000)
declare @alpha table (low nchar(1), up nchar(1))
set @str = 'ALL UPPER CASE and SOME lower ÄÄ ÖÖ ÜÜ ÉÉ ØØ CC ÆÆ'
-- stage the alpha (needs number table)
insert into @alpha
-- A-Z / a-z
select nchar(n+32),
nchar(n)
from dbo.Number
where n between 65 and 90 or
n between 192 and 223
-- append space at start of str
set @str = lower(' ' + @str)
-- upper all lower case chars preceded by space
select @str = replace(@str, ' ' + low, ' ' + up)
from @Alpha
select @str