I'm not so sure. Using the analogy of a computer program, we could think of thoughts either as like the lines of code in the program (in which case they're at one with, or in control of, the processes generating behavior, depending on how you want to look at it) or you could think of thoughts as like the status messages that print "Reticulating splines" or "50% complete" to the screen, in which case they're byproducts of those processes (very specific, unnatural byproducts, to boot).

My view is closer to the latter; they're a way of allowing the brain to make inferences about its own behavior and to communicate those inferences. Opaque processes decide to go to Subway tonight because they've heard it's low calorie, then they produce the verbal sentence "I should go to Subway tonight because it's low calorie", and then when your friend asks you why you went to Subway, you say "Because it's low calorie").

The tendency of thoughts to appear in a conversational phrasing ("I think I'll go to Subway tonight") rather than something like "Dear Broca's Area - Please be informed that we are going to Subway tonight, and adjust your verbal behavior accordingly - yours sincerely, the prefrontal cortex" is a byproduct of their use in conversation, not their internal function.

Right now I'm just asserting that this is a possibility and that it's distinct from thoughts being part of the decision-making structure. I'll try to give some evidence for it later.