Knowledge is useful. If something is known, it can be used as a premise in constructing an answer to a question. But this platitude can quickly lead to scepticism, for there are few things that we know with such certainty that we can safely use them in any inquiry we happen to undertake. I think the way out of this sceptical problem is to adopt a form of pragmatic encroachment - what one knows is sensitive to what inquiries one is undertaking. But if we start with questions, we get a somewhat distinctive form of pragmatic encroachment. It is one that denies any particular link between knowledge and high stakes, as other pragmatic encroachment theories have stressed. And this is good, because the notion of a high stakes choice isn't a very well defined one. But it leads to other problems involving choices between nearly indiscriminable options - such as choosing which jar of Vegemite to grab from the supermarket shelf. I will set out some tricky problems my view faces in these cases, and try to solve them by distinguishing the task of deciding what to do, from the task of figuring out which option has highest expected utility.