I'm with those who consider theory as a source of ideas, but not the final arbiter of them (that being the ear).
A recent example in which a little theory went a long way: A bandmate came to practice with a partially written song. It just alternated between two chords, but it had a cool groove, great lyrics, and a nice vocal melody. He knew it needed more, though, and so he asked the rest of us for suggestions. Another bandmate and I -- the official band theorists -- came up with the same idea simultaneously: Given that the two chords were C#m and D, it occurred to us that they could readily be part of a larger song in the key of A. Response: "Key of A, huh? Interesting idea. I'll fool around with that." A week later the guy shows up with a now-complete song, having added a second part (and lyrics to go with). Now it starts with a verse section that starts on A -- i.e., A - C#m - D -- and then switches to the original C#m/D section for the chorus. And it works just as it "theoretically" should: With the A part firmly established, the C#m/D part now generates a lot of tension -- it sounds like it desperately wants to get back to A -- so when it finally does go back to the A for the next verse, Bam! -- there's a huge resolution/release. (He then built further upon this idea by having the second chorus last twice as many bars as the first time around, thus generating even more tension/suspense as you wait for it to come home to the A.) It's my favorite song in our repertoire now, and I think it's kinda cool that what ultimately made it come together was the simple recognition that the two chords he had started with suggested the possibility of a particular (different) key.
A recent example in which a little theory went a long way: A bandmate came to practice with a partially written song. It just alternated between two chords, but it had a cool groove, great lyrics, and a nice vocal melody. He knew it needed more, though, and so he asked the rest of us for suggestions. Another bandmate and I -- the official band theorists -- came up with the same idea simultaneously: Given that the two chords were C#m and D, it occurred to us that they could readily be part of a larger song in the key of A. Response: "Key of A, huh? Interesting idea. I'll fool around with that." A week later the guy shows up with a now-complete song, having added a second part (and lyrics to go with). Now it starts with a verse section that starts on A -- i.e., A - C#m - D -- and then switches to the original C#m/D section for the chorus. And it works just as it "theoretically" should: With the A part firmly established, the C#m/D part now generates a lot of tension -- it sounds like it desperately wants to get back to A -- so when it finally does go back to the A for the next verse, Bam! -- there's a huge resolution/release. (He then built further upon this idea by having the second chorus last twice as many bars as the first time around, thus generating even more tension/suspense as you wait for it to come home to the A.) It's my favorite song in our repertoire now, and I think it's kinda cool that what ultimately made it come together was the simple recognition that the two chords he had started with suggested the possibility of a particular (different) key.