The combined advances in robotics, mechatronics and artificial intelligence are accelerating the development and production of increasingly efficient and autonomous automated systems. All areas of human activity (industry, production, services, transport, agriculture, health, R&D, military, economic, financial and social) are undergoing or will undergo profound changes as a result of this onslaught.Among the many questions that arise, one that is particularly recurrent when these systems are deployed for free use is that of the adherence of some of these systems (social robots, dynamic interfaces, for example) to those for whom they are intended. It must be said that many of these experiments have been inconclusive and there have been many cases of failure. Using various semiotic hypotheses based on the notion of salience, we intend to show that these failures are due to the difficulty these systems have in identifying the hooks that would allow them to adapt as closely as possible to the environments in which they have to operate. This will provide an opportunity to revisit concepts such as gaps, ellipses, allusions and incorporation as developed by Merleau Ponty and in the cognitive sciences. These concepts, less constrained by structural organisation, seem to us to be better suited to getting as close as possible to the phenomena.