Asymmetric key algorithms, where the key used to unlock the document is different from the key used to lock it, were long considered the "holy grail" for cryptography. With an asymmetric key, I can closely guard the "unlock key" (private key), but shout out the "lock key" (public key) from the rooftops or post it on flyers and anyone could send me documents in secret that only I could unlock. (That unlocks the next, and still current, major problem in that how do you as someone looking to send a secret document to me trust that the person shouting that night from the rooftops or posting that flyer is actually me, much less know which flyers to look for or that I might shout it sometimes from rooftops...)
PGP was the first software built "for the masses" that made asymmetric key crypto accessible to the "common computer users".
(PGP is also notable for attempting to build one such potential trust system to find/trust keys, the Web of Trust model. This is where most of the contemporary arguments about the value of PGP these days are targeted. The Web of Trust is a useful solution, but its a complicated solution that doesn't scale as well as it should and has a bad user experience with a sharp learning curve and a long tail of key management complications for even power users.)
I have a text file. I encrypt it using a key. I send you the encrypted text file. How do you decrypt it? You need the key.
If I can send you the key securely why don't I just send the plain message using that secure mechanism?
Public key cryptography solves that problem.