Automated systems and robotics have been presented for several years as a major stake in the development of companies because they are supposed to increase productivity on the one hand, and improve the health of operators on the other hand, for example by automating difficult tasks. However, as human remains essential to productive systems, the effective collaboration of man and technology depends largely on the acceptability of these new technologies by operators. The assessment of acceptabilité in the design phase of a new technology is therefore an major scientific issue to assess early the costs and benefits perceived by future users. Despite the attention paid to the acceptability of operators, we note that many investment projects related to new technologies encounter difficulties of integration in the socio-professional context of the company, even though Acceptability is good or very good. The purpose of our research is to understand this contradiction and our postulate is as follows: the acceptability assessments are carried out outside the context of use, which does not allow the operators to evaluate concretely the contributions and the limits of the technology. To understand the dimensions of acceptability that can be fairly easily evaluated in an experimental context and those that pose problems we have integrated an experiment around the use of a new robot, during a picking task. The robot automatically follows the operator and carries the packages he has deposited on robot. This experiment took place in 3 stages: - First phase: learning the task of order picking with a manual trolley through the realization of 10 orders; - Second phase: learning how to operate the robot by performing 1 command; - Third phase: realization of 10 commands with the robot, with 7 commands that incorporated into the environment obstacles hindering the operation of the robot. The evaluation of acceptability was carried out following these 3 phases using the UTAUT 2 questionnaire (Venkatesh, Thong and Xu, 2012). To give participants the opportunity to express their inability to answer a question, we changed the value scale by replacing the "neutral" box in the middle with "do not know how to answer" Overall, the results show that the use of the robot leads to an evolution, mostly positive and less frequently negative, of the participants' acceptability, which reinforces the idea of the need to put in a situation of use the operator to better appreciate its future acceptance of the technology. Nevertheless, we also note that some important dimensions of acceptance (Bobillier-Chaumon, 2016) are difficult to assess through our experimentation, particularly the relational and identity dimensions, where people have more frequently meant that they were not able to answer. These results may confirm the interest of experiments in measuring a number of variables, but also the need to associate experimental conditions with conditions and / or methods which can integrate or simulate more satisfactorily relational and identity dimensions of human experience. We consider these as a major factors in the processes of acceptance of new technologies. By replacing the indication "do not know how to answer" with a "neutral" score (e.g. average score), as expected in UTAUT 2, these dimensions are “diluted” in the overall acceptability assessment that seems very good.