The agentic AI introduces a new work organization between employees and virtual assistants. It is also a question of responding to the legitimate concerns of employees feeling threatened in their employment.
Amazon employees are warned. In a letter made public, their CEO, Andy Jassy did not handle the language of wood. The more than a thousand services and applications of generative AI deployed or under construction represent “only a small fraction” of what the e-commerce giant plans to set up. In his eyes, the rise of agentic AI will place the automation of tasks in another dimension. “This should change the way our work is done,” says Andy Jassy. “We will need fewer people to do some work and more people for other types of jobs.” In the end, the Amazon boss expects to reduce his workforce thanks to the productivity gains obtained.
Beyond the particular case that Amazon represents, the rise in power of agentic AI will concern a large number of companies, small and large, with a certain impact on their organization of work. If generative AI chatbots are content to respond to user requests, by prompt intermediaries, they act. Within the limit of a predefined perimeter, they make decisions and perform actions independently with the most limited human supervision. In the case of a marketing campaign, an army of virtual agents can coordinate to automatically generate leads, write impactful content, send the right messages to the right channels, adjust in real time the offer and targeting according to the first lifts of the field and then reassess the prospect scoring.
A learned distribution of work
For Caroline Monfrais, Global Managing Partner, Strategy & Transformation at Wipro Consulting, an AI agent is not a tool but a virtual colleague who occupies a role. “Depending on the process eligible for automation, it is a question of defining these roles and then deciding to entrust them to human collaborators, agents or hybrid teams,” she argues. In a learned distribution of work, agents will be entrusted with the tasks where they are most relevant such as the analysis of large amounts of data or the automation of recurring and time -consuming processes. Humans will focus on missions using creativity, empathy or complex and strategic decision -making.
In a recent press release, Workday spins the metaphor. “While the agents of AI become full -fledged actors”, it is a question of “recruiting them, integrating them, assigning them responsibilities and following their results – exactly as for human collaborators,” said the workflow Automation specialist. The objective is “to define their roles, to control their access to data, to manage their actions and to assess their performance, in order to align each asset – human or not – on the objectives of the company”.
Adecco, user and prescriber
For Pierre Matuchet, please it & digital transformation of the interim Adecco interim (which has chosen the agenforce platform of Salesforce), the agentic AI refers to a concept of increase and not of substitution. “It can increase employees and unload them repetitive tasks so that they can devote themselves to the essentials, namely dialogue with candidates and recruiters,” he said. The first deployment of agents at Adecco relates to the pre-skilling phase, that is to say the prequalification of candidates eligible for a given position. “Before going further, you have to ensure their availability and mobility,” continues Pierre Matuchet. “The agentic AI will generate a game of questions that the recruiter validates before being sent to the candidates.” Currently in the pilot phase in the United Kingdom, the project will be tested in France.
Company Client of Salesforce, the human resources group group was further in its partnership with the CRM leader by creating a joint venture. Called R.Potential, it is the mission, according to Denis Machuel, CEO of the group based in Switzerland, “to help business leaders improve their labor strategies in a new paradigm where human and digital agents collaborate side by side”.
Based on client company data, R. Potential will advise it on the best distribution of tasks between workers and digital agents. It is both a question of managing the resources in real time but also of simulating hybrid work models in the case of peaks of request, mergers-acquisitions or transformation plans. “It is a real technological and organizational rupture, we approach it with great humility,” recognizes Pierre Matuchet.
A new man-machine cohabitation
This new organization of man-machine work is not without creating questions among those concerned. Like workers who have learned to cohabit with factory robots, white passes will have to learn the work shared with software robots. At first, they can legitimately feel threatened in their use or at least in their functional perimeter and their freedom of action.
As part of a change support program, Caroline Monfrais advises to show pedagogy by explaining the assets and limits of agentic AI. “You have to demystify the agentic AI by forgetting the technical jargon. When I show what I use on my computer, my interlocutor is reassured. It is just a conversational interface which very simply allows agents in natural language”, she argues.
For the consultant, not all employees leave with the same level of maturity. “On a project, 15 to 20% of employees, already broken to prompt, are enthusiastic. They concretely receive the profits of automation,” she notes. Xavier Cimino, Senior Managing Director Strategy at Publicis Sapient, confirms: “Employees who have used Copilot for several months have a better understanding of the contribution of AI.”
“Employees must perceive the value they have in gaining the use of agents. And not only for their organization, but also for them,” continues Caroline Monfrais. “The agents of AI will perform all these tasks that we would like to do but that we postpone for lack of time. An employee who spends his day to copy and paste on an Excel file will see his daily life turned upside down. Finally, he should be more fulfilled and motivated.”
An ethical and transparent AI
An agentic AI project requires some prerequisites. He must call on an AI of trust and respond to an ethical and legal framework, complying with AI Acts, labor law and the internal code of conduct of the company. Like an increasing number of companies, Adecco has set up an ethics committee dedicated to AI. Transparency is required. Employees are systematically informed that they work with machines. To avoid the black box effect, the operating mode of AI agents must, moreover, be traceable and explainable.
Involved from the upstream phase, the HR function plays a crucial role in the implementation of AI agents, in particular in terms of work organization and information in the representative bodies of staff. It must assess the social impacts of such a structuring project and anticipate them by implementing training or even retraining actions for the most automatic positions.
The HRD must be even more positioned on the subject what is one of the directions, with marketing, to benefit full of agentic AI contributions. The most common use cases relate to the management of recruitment processes or performance evaluation. As part of a personalized integration course, agents will be able to coordinate to generate and then validate the various tasks that the new collaborator must accomplish when he entered. Publishers like Workday or Cornerstone already offer platforms responding to these issues.
