There is a distinction to be made here between brute utility and morality. If a person commits a crime and serves the time for it and lives beyond the point where that crime has to be reported when applying for a job - should that previous conviction alter their likelihood of getting hired? Society as a whole may decide that it should not, yet the processes which determine the chances of hiring will not reflect the decision made by society as a whole. There will be a very large number of conflicts of this type, on the one hand businesses want to maximise profits and on the other society as a whole wants to impose certain moral and economic values upon those businesses. This is part of a more general tension between different structures within a society. The difference here is that before the conflict was less obvious because the agents involved could hide their motivation, when we are using machine agents for decision making we can easily query them and observe their bias objectively. So rather than creating discord between the various motives of society, it is really only making them more visible. Which I would say is a good thing.