There are, of course, a number of well documented shortcoming with the existing approach (see, for example, McCrae's research: Profiling the Risk Attitudes of Clients by Financial Advisors: The Effects of Framing on Response Validity. I am assuming that we're talking about a robust approach to risk profiling and asset allocation, based on a reasonably accurate "attitude to risk" profiler, incorporating goal specific information and timelines (capacity for risk), thorough analysis of efficient frontiers, and regularly reviewed and adjusted. (Any approach that ignores those factors is bad, without detracting from the underlying principle.)
The main criticism that I encounter is that risk profiling questionnaires are inaccurate. That is obviously true. What you really want to now is what is the customers optimum trade off between risk and return. But given that most customers don't understand return (many don't even understand percentages) or risk (even fewer understand probability, let alone probability distributions), it follows that you have to settle for the closest approximation of this that you can find - which will always be slightly inaccurate.
A number of improved approaches have been suggested, for example:
So it seems to, that despite its failings, there is no real alternative to risk profiling and asset allocation, and the best we can do is to strive to improve our tools for performing the existing process.