I have little to add to Alexander's fine comments other than to reiterate it is perfectly possible to insert UTF-8 encoded perl scalars into MS SQL Server with DBD::ODBC.

Go through his list, make the corrections he suggests and examine the unicode examples in the t/ subdirectory of DBD::ODBC.

In addition to what Alexander said you can avoid that call to decode_utf by adding a filter to your open command e.g., open(FH, "<:encoding(UTF-8)", "file") (see open)

Do upgrade to the latest version of DBD::ODBC if you can as many issues have been fixed since the version you are running.

If you follow this advice and still have a problem try and create a small complete example which does not work, include schema definition, run the code you submit and include the output.