I read somewhere that I can UTF-8 by specifying it at the beginning of the document use utf8;

use utf8; only specifies that the source is UTF-8. If you're reading data from a file, for example, you'll still need to decode that.

open(my $fh, '<:encoding(UTF-8)', $qfn)
    or die("Can't open file \"$qfn\": $!\n");
[download]

Don't forget to encode your output.

s/-/\x{2014}/g; This should turn a hyphen into an em dash correct?

Yes.

\x{2014} works even without use utf8;. It refers to character U+2014, no matter which encoding was used for the source.

The problem is, I only want to do the substitutions on the hyphens which are surrounded by 3 digits on both sides.

The approach you are taking require captures:

s/([0-9]{3})-([0-9]{3})/$1\x{2014}$2/g
[download]

But captures aren't needed here.

s/(?<=[0-9]{3})-(?=[0-9]{3})/\x{2014}/g
[download]

(\d matches some pretty funky stuff in addition to 0-9)

The latter snippet has the advantage of properly handling 123-456-789.