The publication of this new study comes at the right moment, as Commission President Von der Leyen requested in her mission letter to Executive Vice-President-designate for People, Skills and Preparedness Mînzatu “to focus on the impact of digitalisation in the world of work” with “an initiative on algorithmic management”. Trade unions have been calling for such an initiative over the past years, as the use of AI at the workplace has been rapidly increasing with both risks and opportunities for workers. Many workplaces are currently the test ground for new technologies, thanks to insufficient regulation for protecting workers. This must urgently change.

There is no technological determinism. Digitalisation and artificial intelligence (AI) are not black or white for workers. New technologies do not automatically lead to good or bad jobs. The outcome depends on how they are implemented at the workplace, and on whether workers and their representatives are involved in the process from the beginning. This overarching conclusion arises from our new study which takes a deep dive into the impact of digitalisation and AI in industrial sectors. AI is already well deployed in the industry, covering more and uses all cases along value chains, from R&D to after-sales and maintenance. Our study’s focus is on the energy (oil and gas), pharmaceutical, telecom and automotive sectors.

Below are the five key takeaways and recommendations of the study:
1.    AI is not an intangible technology but relies on real infrastructures, energy, water and workers’ labour:
-Racks in data centres (DCs) dedicated to AI require between 6 and 9 times more energy power than classical racks
-Today, DCs consume 2% of world electricity, but the share is set to at least double by 2026 (investments in DCs by Big Tech have tripled in the past 3 years)
-Workers are needed to elaborate algorithms, to “produce, enrich and curate” data, making AI very labour-intensive, especially for the invisible ‘click workers’ outside Europe whose working conditions are appalling 

2.    It is difficult to predict AI’s impact on jobs & employment. AI like any other technology is ambivalent and its impact depends on how it is implemented. But:
- Most at risk are administrative occupations and, unlike previously, white-collars seem more targeted by new AI technologies than blue-collars
-Big risk of polarisation between countries (low-income and high-income), sectors, jobs (with a decline in intermediate occupations and an increase of precarious jobs), and high-skilled and low-skilled workers (in the manufacturing sectors, AI has been associated with the deskilling of workers, especially those with medium qualifications)
-This represents a massive source of anxiety for workers which needs to be addressed through involvement of their representatives

3.    AI might boost productivity, but results are yet too fragmentary to be conclusive:
- Productivity seems to increase among junior or low-skilled workers and for tasks of low complexity
-Effects on productivity depend on how AI is used, including on whether workers are trained to use new tools, and on clear guidance and rules for their usage
-Productivity gains need to be shared with workers (for example through working time reduction; generative AI can substantially reduce the easier routine tasks that ensure a cognitively balanced workload, potentially leading to an excessive volume of mentally strenuous task that cause mental exhaustion)

4.    AI is changing work, its content and the organization of work. Again, its ambivalence can be seen, as it can augment workers performance or threaten their expertise. Here are some of the benefits and risks:
-AI can automate time-consuming, low-value, repetitive or dangerous tasks, enriching work with a higher degree of expertise
-But it can also lead to cognitive fatigue, dehumanisation of practices, weakening of know-how and workers’ autonomy, lack of accountability (unclear responsibility regarding decision-making), surveillance and monitoring
-To reduce the risks and enhance the benefits, it essential to involve workers and their representative in the entire process of AI deployment and usage

5.    Very often workers and their representatives are not aware of AI implementation in their company. There is a strong tendency for companies to skip their legal responsibilities or to comply with them too lately. The current legislative framework (including the AI Act) is insufficient to address the impacts of AI in the world of work and guarantee the involvement of worker representatives. It needs to be supplemented to ensure:
-Involvement as far upstream as possible. Approaches such as social design, would increase the possibility of worker representatives to influence company choices in the interest of workers rights. Experimentation and co-construction should be favoured before implementation and the questionability of algorithm should be guaranteed along the entire process
-Information and consultation processes should take place at the earliest stage and should be adjusted to take into account the regular update of AI systems. Social dialogue and collective bargaining should be ensured to regulate the use of AI to enhance its benefits and eliminate the risks for workers
-The involvement of a third party trusted by both sides should be ensured to guarantee the lack of bias/discrimination and provide necessary expertise.

Isabelle Barthes, industriAll Europe’s Deputy General Secretary said: “Our new AI study comes at the perfect time to provide evidence to the European Commission of the urgency to address the impacts of this technology in the world of work. Workers across Europe have been eagerly waiting for an initiative to tackle AI at the workplace and in particular algorithmic management, so that their rights are respected. We will continue to call for an initiative that ensures the involvement of workers and their representatives in the deployment and usage of AI at the workplace, through early information and consultation, as well as through social dialogue and collective bargaining.”

Read the full study here