There is certainly a limit to the level of detail that can be provided in any explanation. But that can't be an excuse for not responding at all when someone's livelihood is at stake.

There is a more general issue here. As we use more machine learning and AI, the problem of having to explain why a particular decision was taken will come up more often, especially if the decision has grave consequences.

All "AI-first" companies would be well advised to work on this problem. This is like computer security in the 1990s. It's going to be absolutely central to many applications of AI and it will be a key legal issue in future.