There are two senses of "the highest good." In one sense, it refers to that which is always good no matter what and that which is required for all other goods. This is dutifulness. In the other sense, it refers to the best of goods, even if part of that state is only contingently good. The highest good looked at this way combines virtuousness with happiness.
The highest good is the object of pure practical reason, so we cannot utilize the latter unless we believe the former to be achievable. However, in this world, virtue does not necessarily lead to happiness or vice-versa. To aim at one is not to aim at the other, and it depends on chance whether the rest of the world will bridge the gap, rewarding the good. So it seems that pure practical reason cannot apply to us after all.
The flaw in this argument is that it assumes that we exist only phenomenally and thus can be rewarded only here in the phenomenal world. However, on the contrary, we can detect our noumenal existence as autonomous causes. Since we exist in a way other than as we detect ourselves here and now, there may be other times for us to be rewarded.
What happens when practical reason's maxims are connected to theoretical positions about which theoretical reason says nothing? Practical reason simply demanding the object of its desire is not an acceptable reason to believe. Just because the notion of mystic union with God, for example, happens to appeal to me is no reason for me to think that it will happen. But when it is pure practical reason that makes demands, that is a different matter. In that case, the demand is necessary for the faculty of reason as a whole and so commands assent.
The highest good requires the highest level of virtue. This, we can tell by looking inwardly, does not exist in us now, nor is it likely to exist in the foreseeable future. In fact, the only way the fallible human will can turn into the perfect holy will is for it to take an eternity to perfect itself. Therefore, we can postulate that we are immortal. If we fail to make this postulation, either we are led to soften the demands of morality to make them achievable here and now, or we are led to make the absurd demand on ourselves that we must achieve the holy will here and now.
The highest good requires the highest level of happiness as well, to reward the highest level of virtue. We cannot suppose this will come about by chance, even given an infinite amount of time. We need to make the supposition that there is an omniscient, omnipotent God who can order the world justly and reward us for virtue.
So we see that we have reason to believe after all in God, freedom, and immortality, although the first Critique led us to think that these would be unknowable. However, although we are required to postulate all of these, and can in fact detect our freedom by detecting the moral law, we cannot cognize them in the sense of attaining a full intellectual understanding. For truly cognizing something requires that we can sense it, and we cannot sense God, freedom, or immortality.
In fact, it is just as well that we cannot sense God, freedom, and immortality. For if we could truly know these things, awe of God, fear of punishment, and desire for reward would become our strongest motivations. We would superficially act as morality demands, but never from the right spirit of pure duty.
The overall argument for the postulates of pure practical reason requires some examination. We need to get from the claim that the object of pure practical reason is the highest good to the claim that we must suppose whatever is necessary to guarantee the highest good in order to follow pure practical reason. There are two problematic things about this transition. The first is seeing which of two senses of the highest good is one with which we started. The second is making sense of the demand for a guarantee of full satisfaction of the highest good to make aiming at it possible.
Kant is not very explicit about why the highest good is the object of pure practical reason. But we can reconstruct his line of reasoning by looking at his talk of dependence on unconditioneds in the first chapter of the Dialectic. For one to aim at some good state of affairs, like getting a paycheck, one must suppose that there is something that makes it good. In this case, what makes this good is perhaps that it provides one with opportunities for amusement and removes the threat of inconveniences, such as eviction or starvation. We then ask what is so good about being amused or not starving. Ultimately, we are driven back to saying that what one aims at actually is the good state in which the worthy person, oneself, is rewarded, and that this aim is dependent on one's worthiness.
Therefore, if we can somehow reach the conclusion that pure practical reason aims at the good, we have an argument that it aims at one particular good, the good of rewarding the worthy. However, this is not quite either of the "highest goods" as Kant describes them. To reward the worthy does not require that the greatest possible rewarding of the worthy take place, nor is it equivalent to merely bringing worthiness, the precondition for any rewarding to take place.
Even so, if we suppose that the object of pure practical reason is the greatest possible good, why would I need a guarantee of full satisfaction of the object of pure practical reason to be motivated by it at all? To see the issue clearly, consider the analogy of a person who desires to be a great painter. The person may take all sorts of steps designed to increase the odds of this, beginning, of course, with doing some painting. Yet no matter how many art lessons that person takes or how long he or she spends contemplating the past greats of painting, there is no guarantee that the person will become a great painter. He or she increases his or her chances, though, via actions, and also stands an overwhelming chance of coming at least partway to his or her goal. And this can be enough for the person's actions to make sense to him or her.
Similarly, it seems that pure practical reason should be motivating even without a guarantee that its object will be satisfied in full with God's help. Kant does have an answer to this, though. For there must be some reason to think that pure practical reason's use at least makes bringing its object about more probable, if pure practical reason's use is to make sense. But Kant will not accept any contingent or a posteriori justification of the use of pure reason. Even if we could argue that in this world, the goodwill of society rewards the just, Kant would reject this as too empirical to be relevant to ethics and demand his postulates with an a priori guarantee.