My guess is you'll need to decode it at some point too. This all very similar to run length encoding, though I think RLE encodes even single character occurrences. So here's short snippet to encode/decode which you can modify to suit your needs:

use strict;

sub encode {
    s/((.)\2+)/(length $1) . $2/eg;
    $_;
}

sub decode {
    $_ = shift;
    my @list;
    while (/((\d+)?(.))/g) {
        push @list, [$2,$3];
    }
    join '', map { (defined $_->[0]) ? $_->[1] x $_->[0] : $_->[1]; } 
+@list;
}

while (<DATA>) {
    print;
    my $enc = encode($_);
    my $dec = decode($enc);
    print $enc;
    print $dec;
}

__DATA__
XYZAAAAAAAADEFAAcdAA
[download]

Which gives the following output:

XYZAAAAAAAADEFAAcdAA

XYZA8DEFA2cdA2

XYZAAAAAAAADEFAAcdAA
[download]