How should social partners take advantage of artificial intelligence? Franca Salis-Madinier, national secretary of the CFDT Cadres, was the only member of a union organization that was part of the committee on AI that presented its report on March 13 to the President of the Republic.¹. She analyzes for Info-Social RH all the risks to which employees are exposed due to this new technology that brings profound changes. She calls on the Government to fund support for this transformation and for social partners to include AI in social dialogue to control its potential negative effects, but also to fairly distribute productivity gains.

What is the current state of social dialogue on artificial intelligence?

Franca Salis-Madinier: In Germany, agreements on AI have been signed at Deutsche Bahn, the German SNCF, but also in the service and trade sector at H&M, for example with the union Ver.di. An agreement was also closed with AXA in Spain, but not in France. I expressed my astonishment to French business organizations (those interviewed in this report) that a French group was negotiating an agreement on AI in Spain, but not in France, with no response from businessmen.

There are also agreements in the United Kingdom, between the Zurich insurance company, well established in this country, and the Community Union, on a training scheme for workers to support them in new professions, during the introduction of automation, with the aim goal of retaining all employees. In the United States, screenwriters and actors went on strike for four months during the summer of 2023 to reach an agreement with the studios. In Sweden there is an interprofessional agreement concluded in 1987 that allows for discussions to be opened as to how the introduction of a new technology implies restructuring and affects jobs and professions.

In France there are provisions, whether European or French legislative, that could open debates of the same type, but they are not used or are rarely used.

What is the situation of social dialogue in the European Union?

FS-M. : At European level, the social partners signed, on June 22, 2020, a framework agreement on the digital transformation of companies. It establishes that a cyclical and iterative dialogue begins when technologies that may have an impact on employment, professions or that raise ethical or social issues are introduced into the company. AI falls into this framework since it is a technology that evolves periodically and has impacts such as those provided for in the agreement. Therefore, we should start regular discussions on this. Unfortunately, this framework agreement has never been implemented in France. Obviously we want this framework agreement to be translated into a national interprofessional agreement and then applied in the different sectors.

Why wasn't it rejected?

FS-M. : The devil is in the details. The 2021 agreement on teleworking included a small footnote that was introduced by employers and referred to the European framework agreement of June 22, 2020. Business organizations consider this note as a variation at national level and do not want go further. I hope that the presentation of the AI ​​report will make things happen.

“Today the opacity is total because the dissemination and use of these tools are not the subject of any information-consultation with workers' representatives or any discussion about the purposes of their use. »

What risks do AI systems pose for employee rights?

FS-M. : Today, AI tools are distributed without information from employees or their representatives, but can be used to track 360° employee activity. This unlimited surveillance based on permanent data collection is illegal. No one can know whether this permanent data collection will one day be used to evaluate employees or even justify their dismissal.

Today, the opacity is total because the dissemination and use of these tools are not the subject of any information-consultation with workers' representatives or any discussion about the purposes of their use. Some companies hide the distribution of artificial intelligence tools because they fear scaring employees. Since the arrival of ChatGPT in 2022, we are witnessing a new phenomenon that is the use of generative AI tools by employees themselves, but the risks are still there, in particular that of discrimination or quality generated content.

To avoid these risks, it is necessary to evaluate the algorithms and European legislation, the AI ​​Act, provides for the creation of an evaluation system through European and national bodies such as the AI ​​Office. The algorithms used in the work must be audited by independent authorities. It remains to appoint the most qualified supervisory and audit authorities.

What impact will AI have on working conditions?

FS-M. : Beyond issues of surveillance and discrimination, this technology will also affect working conditions and perhaps the health of employees. If AI takes over the most routine and least rewarding tasks, the proportion of activities that require greater concentration and cognitive effort will increase. If working hours remain the same, there is a risk of causing greater stress, especially among managers who are already under high levels of stress. We have surveyed employees in the consulting or insurance sector on this question and they already have this experience and this risk.

How should AI be framed?

FS-M. : In Germany there is a national AI observatory. We proposed the creation in France of a similar structure whose steering committee would include representatives of trade union organizations. There are structures such as LaborIA or France Stratégie that deal with different aspects, but an analysis on the quality and volume of employment is missing, such as the one prepared in 2018 by the Employment Guidance Council in relation to the Villani report.².

In the report presented on March 13 to the President of the Republic, it is also recommended to create specific commissions on AI in the Ministry of Labor and the Ministry of Public Service. With the role of identifying good practices for the introduction of AI in the workplace. The social partners will be closely involved in the management of this work. The report estimates that €5 million should be mobilized over five years to finance studies and observation of the impact of AI on skills, professions and jobs. This report and its recommendations were finalized when a decree³ has reduced certain budgets by 10 billion euros, including that dedicated to work and employment.

This raises a question: do we want to make the deployment of AI an ambition and a priority, and in this case should we finance the support part of this technology, or are we going to oppose money issues to give up the creation of systems that Does it allow us to anticipate these transformations? If we believe we must anticipate and support this technological transformation of jobs and skills, then we must commit this funding. It is absolutely necessary for trade union confederations to establish a system to follow up on the report's recommendations. As part of this report, we have already started conversations with the General Directorate of Labor (DGT) on these issues.

“Artificial intelligence systems can potentially generate very significant productivity gains. One of the main questions is how they will be distributed. »

In which forums should AI issues be discussed and negotiated?

FS-M. : The first place AI should be discussed is the boardroom, where strategy is set and where the purposes of deploying AI systems can be examined. But in many cases these issues are not discussed because the actors involved, whether elected officials or members of management, are not trained. The CSE are the second space for dialogue where these issues can be discussed. And there is a third, these are ad hoc commissions that can be created by company agreement, as provided for in article L. 2232-12 of the Labor Code. The interprofessional and branch field is also a relevant level, because the impact of these technologies will be different depending on the sector.

Are social dialogue actors prepared to participate in technological social dialogue?

FS-M. : Social dialogue actors, whether they are employee representatives or also management, have a very low level of knowledge of AI systems. The choice of these systems and their implementation is left to IT departments and their teams. Therefore, the highest priority is to train social dialogue actors in this technology and its implications. The CFDT Cadres began this approach in 2018 and today we see a very strong increase in the demand for training of our teams.

You have to equip yourself to understand these systems, but the essential thing is the consequences of the uses. Hence the importance of experimenting before choosing these tools. Employees may very well notice that a tool produces incorrect results. Another essential step is to initiate a periodic review of these AI systems that are constantly evolving. This is not a classic technology that always produces the same result.

How could AI help unions?

FS-M. : Trade union representatives especially lack time. To prepare for a negotiation, it would be very interesting to have a summary of the agreements already concluded on a specific topic or sector. The ideal would be to have an AI tool that is trained with all the agreements already published on the Legifrance website. The budget required to build a tool of this type amounts to between 30 and 40 000 euros. Therefore, nothing prohibitive from a financial point of view; It is simply a matter of publishing a viable version of the collective agreements for the training of AI and the DGT should be in the initiative. It is a question of will.

What is special about the current period?

FS-M. : We are living through an important period, because it is the first time that union organizations want and must be participants in a technological transformation of this magnitude. It is an opportunity to renew the framework of social dialogue with actors trained to engage in effective negotiations. AI systems can potentially generate very significant productivity gains. One of the main questions is how they will be distributed. Will companies use it to reduce payrolls or to better train their employees? This will be one of the main challenges of social dialogue.

(1) Report presented on March 13, 2024 by the Artificial Intelligence Committee: AI, our ambition for France

(2) Villani Report presented on March 28, 2018: Making sense of artificial intelligence (AI)

(3) Decree No. 2024-124 of February 21, 2024 canceling credits