Summary Cobots began to be used industrially around 2010 and accidents involving them are still poorly documented. Cobotic risk management is a genuine challenge. The scientific literature shows the existence of various models, methods and tools to manage cobotic risks, all centred around the human operator in the integration of collaborative applications. However, another key human in the implementation of these applications, the integrator, responsible for designing the cobotic cell, is often overlooked. To the best of our knowledge, only two studies dealing with a single software design project to implement cobotic cells focus on the integrator, but solely with respect to feedback about their past experiences with integration. This report differs, in that it concentrates on the integrator’s role in its methodology and utilizes the analysis of the entire integration process as it unfolds. The aim of this report is to identify, in the laboratory, the elements essential to the process of safely integrating cobotic cells, taking into account the variables inherent in the task to be cobotized and in the integrator. To achieve this, the study is broken down into three main steps: (1) characterization of the cobotized industrial tasks and human-cobot interactions, based on visual material from case studies and company visits; (2) laboratory integration of four cobotic cells, i.e., two industrial tasks, each implemented separately by two integrators (each integrator must set up the cobotic cells to perform two industrial tasks); (3) analysis of aspects of the integrators’ decision-making practices for each of the four integration processes. The characterization in step 1 of the project led us to recommend five classes of collaborative applications: (1) direct alternating collaboration; (2) direct assisted collaboration; (3) indirect sequential collaboration; (4) indirect parallel collaboration; (5) occasional space-sharing without collaboration. The definition of these classes is useful to any integrator wishing to initiate a risk analysis of a cobotic installation. The risk analysis begins by determining the limits of the system to be installed, as defined by the ISO 10218 robot and robotic devices standard and, more generally, as defined by the ISO 12100 safety of machinery standard. In light of the results of the three steps of the study, this report recommends a tool for determining the limits of a cobotic installation. These limits are the variables inherent in the task to be cobotized, which were noted as the integrations and various stages of the methodology rolled out. We found that, of all the variables influencing the four integration processes studied, the first three elements listed below (linked to the task to be cobotized), and the last two elements (associated with the integrator), were essential in these processes: (1) the choice of cobot; (2) the type of part to be handled and the type of robotic tool; (3) cycle time and productivity constraints; (4) the integrator’s training in machinery safety in general, and in cobotic safety more specifically; (5) the information related to safety or productivity that integrators receive from those around them, as this prompts them to question the initial choices made and correct them if necessary (feedback). At the end of the report, we offer some food for thought on these variables.