G'day stray_tachyon,

There's a number of things you can do to improve this code. Here's a list of some of the things I found.

• Always use strict and warnings. See "perlintro -- Perl introduction for beginners" for further discussion.
• Use lexical variables, not package variables. Use them in the smallest possible scope. They're faster and you won't suffer from any of the classic problems associated with global variables.
• Order your list of unions. If you have some domain-specific knowledge regarding which are more likely to be found, put them first; if not, put the shortest strings first because it's quicker to match a short string than a longer one.
• Canonicalise your data. It's quite possible there's extraneous whitespace in the union names; long names may be spread over two lines in the agreements; someone may have failed to capitalise part of a name. Whatever you do to the agreement text, do the same to the list of unions.
• Avoid regexes wherever possible. Perl's string handling functions, such as index, are typically, measurably faster than a regex.
• If you're only looking for the first match, don't use grep; List::Util's first function is likely a much better choice.
• Given you're dealing with .txt and .pdf files, these are probably relatively small and would easily fit into memory. Slurp the entire text of each file into a single string and perform your matches on that. (See the special variable $/.) If any names are spread over multiple lines, attempting to match on a line-by-line basis will be pointless.
• Check your work as you go. Use print statements to check that variables contain reasonable values. Your regex /.pdf.txt/ is quite clearly not doing what you want: a print statement showing that the array @files was empty would have told you that.

I put together a short script to show how all of those points might be implemented. I also dummied up some highly contrived data just to give the script something to work on.

Here's union.txt (note the second line has an extra space):

$ cat union.txt 
ABC Union
XYZ  Union
[download]

I then created a number of very short files in two directories. These have one, two or no matches; one has a name spread over two lines with a slew of extra whitespace; one file is completely empty.

$ for i in agreements other; do for j in `ls $i`; do echo "*** $i/$j *
+**"; cat $i/$j; done; done
*** agreements/abc.txt ***
...
ABC Union
...
*** agreements/abc_xyz.pdf ***
...
XYZ Union and ABC Union
...
*** agreements/def.txt ***
...
DEF Union
...
*** agreements/pqrpdf ***
... temp data ...
*** agreements/xyz.pdf ***
.................. XYZ
   Union .............
*** other/dummy_empty ***
[download]

Here's the script to process that data:

#!/usr/bin/env perl

use strict;
use warnings;
use autodie;

use File::Spec;
use List::Util 'first';

{
    my $union_file = 'union.txt';
    my @dirs = qw{agreements other};

    my $unions = get_unions($union_file);

    print "Unions to check:\n";
    print "\t$_\n" for @$unions;

    process_files($_, $unions) for @dirs;
}

sub get_unions {
    my ($union_file) = @_;

    open my $fh, '<', $union_file;

    my @unions;

    while (<$fh>) {
        chomp;
        y/ / /s;
        push @unions, $_; 
    }

    return [ sort { length $a <=> length $b } @unions ];
}

sub process_files {
    my ($dir, $unions) = @_;

    print "Prcessing directory: $dir\n";

    opendir(my $dh, $dir);

    for (grep /\.(?:txt|pdf)\z/, readdir $dh) {
        my $path = File::Spec::->catfile($dir, $_);

        print "\tProcessing path: $path\n";

        my $text = do { open my $fh, '<:crlf', $path; local $/; <$fh> 
+};
        $text =~ y/ \n/ /s;
        my $found = first { -1 < index $text, $_ } @$unions;

        if (defined $found) {
            print "\t\tMATCH: $found\n";
        }
        else {
            print "\t\tNo matches found.\n";
        }
    }

    return;
}
[download]

Here's the output:

Unions to check:
    ABC Union
    XYZ Union
Prcessing directory: agreements
    Processing path: agreements/abc.txt
        MATCH: ABC Union
    Processing path: agreements/abc_xyz.pdf
        MATCH: ABC Union
    Processing path: agreements/def.txt
        No matches found.
    Processing path: agreements/xyz.pdf
        MATCH: XYZ Union
Prcessing directory: other
[download]

Take whatever ideas, or actual code, you want from that. I'd recommend you run Benchmarks to see what improvements you're making: probably also useful for the person who told you "... my codes aren't efficient ...".