The development of systems based on Artificial Intelligence (AI) capable of collaborating effectively with humans is a crucial research topic in contemporary societies, where the decision-making processes are progressively being delegated to algorithms. The conception of AI as free from the flaws of human reasoning and capable of arriving at objective and neutral decisions is false from an epistemological point of view. Besides, it is dangerous from a practical point of view too as it pushes organizations and work teams towards algocratic forms of governance that impoverish the cognitive dimension of such organizations. Based on the study of the conception, design and realization of the robot Maxwell, the article attempts to outline a design philosophy for collaborative intelligent machines that abandons the myth of the infallibility of AI and develops a perspective of true integration between man and machine. This perspective is tackled both from a technical-scientific point of view, relating to how to create machines capable of interacting with humans, and a socio-organizational one, relating to the kind of impact these machines can have on human work.
