In May 2025, Ceemet, as coordinator of the EU‑co‑funded AI@WORK project, launched an analysis on the impact of Artificial Intelligence on labour markets, company management, and workforce planning. The initiative aims to help employers and workers understand the risks and opportunities associated with AI, and to support companies in adopting these technologies responsibly. With the involvement of the German Economic Institute (IW), Institute of Applied Industrial Engineering and Ergonomics (IFAA) and the support of the European Employers’ Institute (EEI), the first factsheets and business cases started to be published in September 2025. As the project reaches its midpoint ahead of its planned conclusion in April 2027, Ceemet is sharing preliminary findings and emerging conclusions.
AI regulation in the EU and its impact on the workplace
The evolving EU regulatory framework places new responsibilities on companies using AI. A first essential step is determining whether a company is acting as an AI deployer, using systems purchased from external providers, or as an AI provider, developing and marketing its own systems. This distinction shapes the obligations that follow.
For companies deploying AI, compliance requires strict adherence to the provider’s instructions, ensuring that human oversight remains central in the Human resources field, particularly in reviewing and controlling input data. Workers and applicants must be informed when AI tools are used in processes that affect them, and system logs must be preserved to guarantee accountability. Companies must also monitor the performance of AI systems and report incidents to both the provider and the relevant supervisory authorities. As outlined in the EU framework, these obligations already exist and cover a wide range of risks that could apply in the use of AI. This initial factsheet confirms that the EU framework provides satisfactory coverage and clarifies the distinctions between providers and deployers, which had previously been ambiguous in the AI Act.
Moreover, this factsheet reaffirms the decisive role of the social dialogue in the successful introduction of AI. Open communication, early involvement of employee representatives, and clear explanations about how AI systems function help reduce concerns and build trust. In several EU countries, informing works councils about planned AI deployments is not only good practice but also a legal requirement.
Following these legal considerations, it is important to note that the potential fears and risks associated with AI, as perceived by employees, are often overstated. Furthermore, there is also a tendency to overestimate the extent to which employers are currently using AI.
AI and workforce planning: the insights from the Airbus Case
One of the project’s key business cases in the Metal, Engineering and Technology industries (MET) examines how Airbus is using AI to support workforce planning. Faced with a slowdown in external hiring, the company needed to strengthen internal mobility. AI now helps identify open positions that match employees’ skills and mobility preferences, supporting retention by offering internal career opportunities to workers who might otherwise look at other opportunities after upskilling.
The introduction of AI at Airbus followed a structured approach. A dedicated HR team, known as “AI ambassadors,” was created to explore solutions. Business and digital teams collaborated closely, and the system was deployed across several countries. A formal agreement with social partners addressed concerns related to data protection and security, supported by a comprehensive dossier developed by Airbus.
The company views current AI use as a temporary support tool rather than a replacement for HR expertise. The main challenge lies in adoption, requiring significant upskilling across the HR team. Managers must also be supported to overcome barriers and embrace new ways of working.
The benefits of this type of AI application are already visible. Matching processes are faster and more accurate, talent retention is strengthened, and leadership pipelines are reinforced. This practice of AI does not replace human judgment but enriches it by connecting competencies and experience in ways that support better‑informed decisions.
A key lesson from Airbus is the importance of early engagement with employee representatives. Initially underestimated, this step proved essential for the acceptance of AI tools. Looking ahead, Airbus is testing how AI can support headcount forecasting for future operational plans.
AI and competitiveness: the lessons of an SME
The second business case in the MET industries highlights how a medium‑sized industrial company, Heuille & Fils, is using AI to enhance competitiveness. Initially unsure how AI could apply to their operations, the company quickly realised that traditional tools such as Excel were no longer sufficient for quality management. Within two months, they developed a prototype platform integrating strategic and operational planning, supported by an AI‑enabled data model.
Today, the company uses generative AI in areas such as predictive maintenance, procurement, production, risk analysis, sales, and training. When used appropriately, AI reduces rather than increases workload by automating low‑value tasks and allowing employees to focus on decision‑making and value creation.
Introducing AI into the workplace was facilitated by the fact that MET industries are already highly automated. Employees welcomed the automation of planning processes, especially those who had previously performed these tasks manually. Once workers began using AI tools, new ideas emerged naturally, for example, for defining customer requirements, conducting risk analyses or structuring next steps. However, fully automated risk analysis and self-correcting processes remain prohibited, and human supervision is mandatory. This is especially important given that no business leader would ever allow AI to make decisions on their behalf. They remain responsible for the quality of their products, which is particularly critical in fields such as aeronautics, where a single error could cost hundreds of lives. That is why AI tools are accessible to nearly everyone in the company, but full automation is not yet considered realistic.
The benefits of AI use are already tangible in the company. AI helps identify operational risks, mitigation measures, appropriate equipment, and procedures. It reduces time spent on monotonous tasks and improves overall efficiency. Employees appreciate having an assistant that supports their work without replacing their expertise.
The experience of Heuille & Fils, with 20 employees, offers broader reflections. AI capabilities are often overestimated, and end‑to‑end automation remains out of reach for many industrial processes. AI does not make decisions; rather, it supports analysis and insight generation. At the same time, the rapid evolution of AI introduces a new challenge: the pace of change itself.
Digitalisation for competitiveness, the case of Miele
Miele has started to develop AI tools for predictive maintenance[1], and Miele is also exploring the use of AI in quality management, improving material utilisation and minimising waste, as well as in quality inspection. Miele is also using generative AI in a pilot project to evaluate employee knowledge. In this project, employees describe their activities in natural language, and the AI assistant automatically structures and formalises the information. For the employees, “AI is like an old colleague who provides assistance when they encounter difficulties”, explained Dr Jonas Osterloff, researcher for the Data Driven Optimisation for Miele.
Miele’s AI journey has shifted from technology-driven experimentation to a clear, problem-oriented approach, supported by a scalable data platform. The company is working to standardise its facilities to record their energy use and productivity, as well as facilitating new use cases such as predictive maintenance. While office functions embrace new technologies, there is still major potential on the shop floor.
Regarding AI adoption, Miele actively involved employees, recognising that even the best application will not help if it is not used. The workers in the press shop had to train the AI, which required intensive communication to address their fears about losing expertise. The use of AI tools for predictive maintenance has shown that, at times, human intuition alone can outperform algorithms, thus proving that technology and people must evolve together.
The benefits of AI use are yet to be determined as the company is currently assessing the realistically achievable gains while accounting for the initial investment costs. Miele noted that they sometimes develop hypotheses about the feasibility of a solution, but that its effectiveness usually cannot be confirmed at the beginning.
At that stage, Miele has not yet achieved an overall return on investment, although this may be the case in some individual use cases. The impact of AI requires careful monitoring because models only work reliably under the conditions under which they were trained. Introducing AI, therefore, requires new, transparent co‑determination processes with the works council, since traditional rules are not well-suited to systems that evolve over time. Workers’ representatives need solid training, as adapting these processes is both challenging and essential for responsible participation in AI deployment.
Mid‑term reflections from the AI@WORK project
The various cases and analyses that have been examined thus far have repeatedly highlighted several key findings. First, it is vital to emphasise that human oversight remains essential for both compliance and trust. AI is most effective when applied selectively, supporting human expertise rather than replacing it. Secondly, social dialogue is a key to successful AI deployment. It allows for reducing potential fears and helps to better identify the needs of the company and its employees. Finally, the rapid pace of technological change in itself is becoming a challenge, requiring continuous learning and a high capacity for adaptation.
As the project enters its next phase, the focus will be on the influence of AI on business competitiveness, skills needs, and the emergence of new skills. This will be accompanied by a survey to understand how businesses are responding to AI, as well as the targeted activities and workplace changes this is bringing about.
—
[1] Predictive maintenance (PdM) is a proactive strategy that uses artificial intelligence and machine learning to analyse equipment data in real-time and predict failures before they occur.