I would try to cover the following topics as they seem to be appropriate to your cow-orkers, together with the problems. I would try to structure the path to these problems by asking the cow-orkers the "right" questions leading them to the problem domain ("What task do you do when loading the new customer data" could lead to DBI or to SQL::Statement or to Template::Toolkit, depending on whether you (want to) load your data through DBI or the datbase loader tool for example).

Topics to know would be:

• Munging data:
  s///, m//, Text::CSV, SpreadSheet::ParseExcel, SpreadSheet::WriteExcel and possibly automating Microsoft Office if your platform has OLE)the "Unix shell to Perl traveler's handbook", UNIX 'command' equivalents in Perl
• Launching subprocesses (or, "how to use Perl as if it were a shell script")
  perlipc, backticks, sed, awk. A good example might be how to transparently read from a possibly compressed file, by reading from open my $fh, "gzcat $file |"; or somesuch.
• Something for the build process could be a reprise of perlipc, possibly together with Expect for automating some user inputs or Net::Telnet or Net::SSH for kicking off processes remotely, or even parallelizing stuff via Parallel::ForkManager. Or maybe transferring files via Net::FTP.
• Things where Perl "shines" over "The Shell" are file/terminal IO, arrays and dictionaries, while Perl is not necessarily as convenient for oneliners and process control as some shells. I'm sure your shell-loving cow-orkers will find plenty of things where Perl is uglier than shell :)