Can't blame the bean counters, Krusty. They are just doing their job. It is up to the flight operations departments to say "no" to the bean counters.
I blame the airworthiness authorities for being so slow to recognise that there are fatal flaws with the man/machine interface in these highly automated aircraft. These problems were evident well before Air France 447, and are not confined to Airbus types.
I find it frightening just how overwhelmed a crew can become when faced with even minor problems in these aircraft. These are highly trained and experienced crews, but their effectiveness is reduced to the point where a couple of passengers at the controls would be just as effective.
The interface seems to be designed primarily around normal operations (where all systems and instrumentation are functioning properly). Given that these aircraft spend 99.95% of their time in such normal operations, this is understandable, but for abnormal situations such as those AF447 found itself in, the interface can then work against the pilot - as we have seen.
What seems to have been lacking in the certification process is some rigorous testing of the human factors when things go abnormal. At the moment, this testing is being done by real world pilots on real world flights, with passengers on board.
Every now and then, we get some of those test results back. AF447 for example.