FES PARIS COLLECTIVE BARGAINING PRACTICES ON ARTIFICIAL INTELLIGENCE AND ALGORITHMIC MANAGEMENT IN EUROPEAN SERVICES SECTORS Marta Kahancová December 2024 CONTEXT Artificial intelligence(AI) refers to machine-based systems capable of making predictions, recommendations, or decisions with minimal human input or oversight. The definition of AI can be simplified to include algorithmic management tools to enable automated or semi-automated decision-making concerning workers and their surveillance. The use of AI at the workplace gives rise to concerns about data protection, privacy, power relations, and human rights. To address the challenges arising from the growing use of AI in people management, a 2024 study commissioned by the Friedrich Ebert Stiftung’s Competence Centre on the Future of Work and UNI Europa explored trade union preferences and the current state of collective bargaining regarding employers’ use of AI-related tools in relation to workers in the European services sectors. The analysis draws on twofold original evidence: – an online survey of 148 trade union representatives affiliated with UNI Europa across 32 countries, – analysis of 31 collective agreements that include provisions concerning the use of AI. The survey results reflect trade unions’ current experiences and preferences, general opinions on bargaining over AI-related challenges, expected union actions to advance bargaining on this theme, and examples of good practices regarding AI-related clauses in collective agreements. KEY FINDINGS Bargaining on AI issues is emerging but is not yet as widespread as bargaining on other aspects of working conditions. Among 90 survey responses, only 20% of trade unions reported having a collective agreement addressing AI-related issues at the organisation or sector level. This means the majority of trade unions(69%) lack collective bargaining agreements on AI, while 11% are unaware of the existence of such agreements. If collective agreements contain stipulations on the use of AI at the workplace, these refer mostly to(a) employee training on new AI tools including the risk related to the usage of AI; (b) union involvement in the introduction of new technologies; and(c) working time regulation. While most of the analysed collective agreements contain only general references to the use of technology, several agreements in Italy, Germany, Norway, and Spain serve as examples of more detailed rules and arrangements. These address topics like the right to disconnect, digital rights of workers in the workplace, information-sharing, and business control. 1

Percentage of agreements analysed referring to specific topics Training for employees and/or management on new Al tools (including the risks related to Al usage) Employee/trade union involvement when new technologies are introduced Impact of Al/AM systems on working time and the right to disconnect Other topics Respecting the relevant privacy/data protection legislation Employee/trade union involvement in data protection Not specified 10,34% (3 CBAs) Al/AM tools used for monitoring and worker surveillance (e.g., software/devices to track physical or digital worker activity) 10,34% (3 CBAs) Use of AI/AM in recruitment process, work organization and worker’s assessment 3,45% (1 CBAs) 0 10 20 30 75,86% (22 CBAs) 62,07% (18 CBAs) 48,28% (14 CBAs) 48,28% (14 CBAs) 41,38% (12 CBAs) 37,93% (11 CBAs) 40 50 60 70 80 90 100 100%= all CBAs that refer to the introduction of Al and/or AM at work Source: WageIndicator Collective Agreements Database, accessed in November 2023. N=30 SAMPLE COLLECTIVE AGREEMENT WITH AIRELATED STIPULATIONS The collective agreement of IBM on the introduction of AI systems, concluded in 2020 in Germany(IBM Konzernbetriebsvereinbarung über die Einführung und den Einsatz von Systemen der Künstlichen Intelligenz), can serve as a sample agreement regarding the scope of regulating AI in relation to people management. The agreement stipulates that the use of AI should improve decisions taken at the company, but in no way should it replace human decisions. It defines the extent of damage that the use of AI can cause to workers and categorises various types of AI tools into groups from no damage to high damage. It also categorises the level of transparency of AI-related tools and which groups of workers are affected by the regulatory impact of these stipulations. Furthermore, the agreement establishes the introduction of an AI ethics board, which is responsible for overseeing compliance with and the further development of the agreement. The full text of the agreement in original language can be consulted here: https://wageindicator.de/arbeitsrecht/datenbank-dertarifvertrage/konzernbetriebsvereinbarung-uber-dieeinfuhrung-und-den-einsatz-von-systemen-der-kunstlichenintelligenz-artificial-intelligence. FUTURE EXPECTATIONS AND PRIORITY ISSUES FOR TRADE UNIONS Collective bargaining on AI-related matters is expected to gain greater importance. In 2023, 42% of UNI Europa affiliates participating in the survey were already engaged in discussions and negotiations on various AI-related topics, even though these discussions may not constitute collective bargaining in a strict sense. Priority issues for unions include: – workers’ right to challenge decisions made by automated systems and to receive advice from external data experts – data protection, worker privacy, the effects of AI on working hours, the monitoring of worker activities, automated scheduling of work shifts. Unions also strongly emphasise the need for the right to information and consultation concerning the use and evaluation of AI tools, and the demand for training for staff and management on using AI tools. Themes in which unions are least interested in collective bargaining include access to workers’ emails, voice analysis or the use of chatbots instead of humans. 2

RECOMMENDATIONS FOR TRADE UNIONS To strengthen union expertise in bargaining on AI-related issues, recommendations based on the study suggest that unions should increasingly familiarise themselves with agreements that already contain AI-related stipulations (11% of surveyed unionists were unaware of such agreements, and 69% of surveyed unionists did not have a collective agreement containing AI-related stipulations). Unions can seek inspiration in cross-sector and cross-national cooperation to identify sample agreements that can be adapted to their specific needs. Union action can be twofold: – with focus at the company level – tailoring specific collective agreements to the need of workers and their workplace; – with a policy focus at the national and the European level – seeking regulation on the transparency of using AI-related tools in people management, assessing worker risk and a fair and transparent co-existence of algorithmic and human decisions with limiting the potential damage for employee rights. This target can be reached, i.e., via cross-border union cooperation and thematisation of the AI Act, as well as joining forces in demanding a Directive regulating the use of AI at workplaces. REPORT ACCESS AND REFERENCE Collective bargaining practices on AI and algorithmic management in European services sectors Simona Brunnerová, Daniela Ceccon, Barbora Holubová, Marta Kahancová, Katarína Lukáčová, Gabriele Medas. Brussels : Friedrich Ebert Stiftung, Competence Centre on the Future of Work, 2024. – 29 Seiten= 3,5 MB PDFFile. –(New technologies at the workplace) Electronic ed.: Brussels : FES, 2024 ISBN 978-3-98628-550-0 https://library.fes.de/pdf-files/ bueros/bruessel/21074.pdf CONTACT Fondation Friedrich-Ebert 41 bis, bd. de la Tour-Maubourg 75007| Paris| France Tel.+33(0)1 45 55 09 96 Fax:+33(0)1 45 55 85 62 https://paris.fes.de info.france@fes.de 3