Instead of thinking ”Well, only 3% of our users are harmed by this solution so that’s not too bad.”, think this:
”What can we learn from the human beings representing those 3% in order to shield them from harm, and involve them to help boost the well-being of all users?”