In one of his very early articles "Two Concepts of Rules" Rawls describes the problem, common both to utilitarians and their critics, of mixing up two quite distinct conceptions of rules. On one view of rules, which Rawls termed the "summary" view, rules are looked at as summaries of past decisions arrived at by direct application of a general principle (such as the principle of utility). If you take this conception as an accurate account of rules then you view them as arising from inductive generalisation from particular cases and take general rules then as merely something that sums up the cases. If this view is right then it also follows that each one of us can keep interrogating the correctness of rules and whether it is right to follow them in a given instance. The rule is really applied case by case; although you do get general rules you never take them as settled but as in constant need of possible revision.
Utilitarians themselves often speak of rules in accord with the summary conception and critics of utilitarianism standardly assume that this conception is the right way of seeing the utilitarian view of rules. However, Rawls suggests in this paper a different way of viewing rules, one to which utilitarians would be as capable of appealing as anyone else. This alternative view of rules is termed the "practice conception". On this view rules define a practice and only make sense within the scope of the said practice. For a practice to exist there must be public acknowledgement of the rules in question and general understanding of the mode of application of them. It would be the practice itself that would be the basis of appraisal (utilitarian or otherwise) and not the rules internal to it. On this basis it would be false to say that utilitarian criteria allow one to, for example, break a promise if this is viewed as having the best overall outcome since to take such an attitude is to violate the basic standards internal to the practice. It may be that promising, as a practice, is not good for furthering overall welfare but then it is the practice as such that would be subject to such evaluation and not some instance or particular case.
On the practice conception the rules are prior to the cases that fall under them and so are not inductive generalisations. Further, individuals don't have general authority to continually question the practices and appeal to ways of reforming them. Particular cases wouldn't be exceptions to the rules of the practice but, instead, means of qualifying or specifying the rule.
It is true that if utilitarians are viewed as having a practice conception rather than a summary one then the response to them will have to be much more subtle than is common. Leaving utilitarians aside, however, it is worth asking whether this distinction is not required for other ethical theories. It is clear that a lot of objections to the categorical imperative are similar to objections to utilitarianism in viewing Kant as holding a summary conception that requires the direct application of the moral law to the particular case. Hegel's notorious "empty formalism" objection is certainly of this type since it assumes that Kant has no right to think through the internal features of a practice. However the practice conception fits the actual use Kant makes of the categorical imperative much better than the summary conception.
Nor is it just the case that the distinction Rawls proposes here can be extended beyond the particular example of utilitarianism to other moral theories. It also seems a better example of how philosophical reflection generally should go on the nature of rules. If rules are viewed as arising from a process of inductive generalisation then one is stuck with a simple empiricist account of laws. It is central to the objection to such a view that laws create objects rather than simply "mirroring" them. It is through seeing objects as the product of establishment that we can view them as altering through the formation of understanding of principles. Reform would not be of particulars but of the manner of grasping and ordering them.