Using aitap's answer for finding the 'b's in Company_A, (I think the assignment to '()' isn't necessary), I reproduced your output with this code.

#!/usr/bin/perl
use strict;
use warnings;
use 5.014;
use Data::Dumper;

my %hash = ( Company_A => { 
                             Identifier_1 => ['a', 'b'],
                                Identifier_2 => ['b', 'g'],
                                 Identifier_3 => ['c']
                             }, 

             Company_B => { Identifier_1 => ['a'],
                                 Identifier_2 => ['g'],
                                 Identifier_3 => ['x']
                             });

my $b_count = map { grep {$_ eq 'b'} @$_ } values %{ $hash{Company_A} 
+};

for my $company (sort keys %hash) {
    my (@ones, @multi);
    
    my $href = $hash{$company};
    for my $id (sort keys %$href) {
        my $aref = $href->{$id};
        if (1 == @$aref) {
            push @ones, $aref->[0];    
        }
        else {
            push @multi, $aref;    
        }
    }
    if (@multi) {
        for my $m (@multi) {
            print "$company ", join(",", @$m, @ones), "\n";
        }
    }
    else {
        print "$company ", join(",", @ones), "\n";
    }
}
[download]

C:\Old_Data\perlp>perl t6.pl
Company_A a,b,c
Company_A b,g,c
Company_B a,g,x
[download]