It sounds a bit circular to say that the units of selection are whatever the environment is selecting for (and against). But we have no choice; fitness functions are highly complex, multidimensional, and change over time. One cannot say a priori what the system is selecting for; one can only observe selection retrospectively, and thus only take an empirical, backward-looking approach to defining units of selection. (The Origins of Wealth, 283)One can predict the outcomes of simple systems, but complex systems are inherently unpredictable. We cannot predict how individuals will interact to create the environment which selects for or against the units of selection. As Hayek observed, we can perhaps make pattern predictions, but that is the best we will ever be able to do. Similar kinds of things happen when similar interactions take place in similar environments.
Because of this, all entrepreneurial activity is a kind of "deductive-tinkering" (286). It is deductive because we have some idea of what works or may work in the environment we are working in, because we have seen what works in the past in similar kinds of environments; it is tinkering, though, because we can never know for certain if what we're doing will in fact work. There may be some small thing in the environment which will have an unepxected effect, and the entire system can collapse -- meaning your plans will fail.
Beinhocker is primarily talking about selection of business plans in a given economy at a given time, but the rule is good for any complex evolving process. It is true of social technologies. It is true of art forms -- there may be works which are ahead of their time in a real sense, meaning they will fail now, but may find their audience in the future. The most successful innovations are always those on the borderlands of the known and the unknown, blazing a trail for the future. However, just because something is well-known doesn't mean it will survive in the new environment. We may be able to appreciate Medieval romances, but I doubt if one were to try to write one now that it would succeed -- or that many would put up with the way women are portrayed. The same is true of business plans and social technologies.
The bottom line is that we are seeing several co-evolving complex processes at work. Business plans are always evolving, in relation to the evolution of physical technologies and social technologies. And all of this is taking place on the platform of interacting embodied minds creating complex, emergent social networks which in turn affect the complexity of the brains of the agents in the social systems, which affects social interactions that in turn makes the social environment more complex. The embodied brain and its environment (the various social orders and their economies) co-evolve to ever-more complex network structures. Out of this environment new social technologies emerge, new physical technologies emerge -- both then contribute to social complexity -- and all of this affects the kinds of business plans, which are made of evolving modules undergoing selection, that will work.
Now who in their right mind thinks they can predict the outcome of all of this?