Michael Dummett writes (page 195)
Once the justification of deductive inference is perceived as philosophically problematic at all, the temptation to which most philosophers succumb is to offer too strong a justification: to say, for instance, that when we recognize the premisses of a valid inference as true, we have thereby already recognized the truth of the conclusion. If that were correct, all that deductive inference would accomplish would be to render explicit knowledge that we already possessed: mathematics would be merely a matter of getting things down on paper, since, as soon as we had acknowledged the truth of the axioms of a mathematical theory, we should thereby know all the theorems. Obviously, this is nonsense: deductive reasoning has here been justified at the expense of its power to extend our knowledge and hence of any genuine utility.
I assume that the temptation he is referring to is to believe that all we have to do is verify the premises to accept a validly derived conclusion. We've justified the deductive inference so strongly that we somehow lose the power of the deductive inference to "extend our knowledge".
I can see how knowledge is not "getting things down on paper". We have to know what is on that paper for the data written there to become knowledge. However, I think he is referring to something besides knowing what is written down that we should be doing so that deductive inference can extend our knowledge. Perhaps it is to understand the proof as well? Or is Dummett referring to something else?
Dummett, M. (1991). The logical basis of metaphysics. Harvard university press.