Today it's awk(1)
I use it almost daily—from simple column-extraction (specifying column separators and mashing together various columns feels easier with awk than with sort(1)), to summing and running totals, to aggregating counts of data, to reformatting text, etc…so many little uses pop up.
It's available on every POSIX platform making it easy to write cross-platform utilities without having to install additional run-times like Python/Ruby/Node and the heavy dependencies that come with them.
I've even written cgi-bin/ scripts in awk, allowing dynamic data processing on my stock OpenBSD systems with httpd+SlowCGI without non-stock software in the chroot.