Advances in the way humans and robots works side by side represent one of the untapped potential to get the best from machine efficiency in combination with human ingenuity. To keep up with the rapid evolution foreseen in robotics and realize the robot’s potential as invaluable assistance to humans we argue that today’s practices for ensuring safe and reliable human-robot interactions (HRI) need to be rethought. We argue that systems with a safety/design philosophy primarily based on preventing humans from doing errors will not be able to capitalize on the rapid evolution of HRI and can pose constraints that limit the performance of the complex socio-technical systems they are imposed upon. To improvise or to make errors should be considered a natural part of human nature and our systems and organizations should be designed to be resilient in the face of deviations, errors and unexpected events.