It is not hide and seek but seek not to hide: Shadow AI, Job Crafting and the Power of Social Dialogue
By Pierrette Howayeck, Postdoctoral Researcher at ESSCA School of Management
This blog is produced as a part of the Reshaping Work Fellowship Programme. The opinions and views expressed in this publication are those of the author. They do not purport to reflect the opinions or views of Reshaping Work or organisations that have supported the programme.
One critical question is often overlooked: How will AI change the way workers experience their jobs?
Work is not just a bundle of tasks; in fact it is something people actively shape. Research on job crafting, the process by which workers actively reshape their jobs by changing tasks, relationships, and how they view their role’s purpose, shows that employees constantly redesign their roles by adjusting tasks, building relationships, and reframing their purpose in order to make work more meaningful. If AI is truly going to reshape work, it will reshape this space for job crafting as well.
AI as a tool for crafting work
At its best, AI can be an ally in job crafting. Teachers use generative AI to prepare lesson plans faster, leaving more time to interact with students. Consultants use AI to draft initial project concepts and frameworks, freeing them to focus on strategic thinking and client relationships. Leaders employ AI to optimize the tone and clarity of employee feedback, allowing them to spend more time on meaningful coaching conversations. Knowledge workers use AI to transform lengthy emails and documents into actionable to-do lists, streamlining their workflow and priorities. In these cases, AI removes the mundane to allow workers to craft their roles around what they find most valuable.
When workers have autonomy over how they integrate AI into their roles, the technology can inspire new ways of working. This bottom-up innovation often produces the most meaningful and sustainable changes. However, this potential is only realized when people are free to experiment and adapt.
Shadow AI: bottom-up job crafting in action
Despite organizational enthusiasm for AI, formal implementation often fails to deliver on its promises. Recent research by Inria and datacraft found that despite a decade of AI investments in organizations, only 20% of “proof of concept” projects are industrialized. Even among those projects that advance further toward pilot , MIT research shows that 95% of generative AI pilot programs fail because companies avoid the organizational changes necessary to make AI work effectively.
This gap between corporate AI ambitions and workplace reality has led workers to find their own solutions. Workers across sectors are now using ChatGPT, MidJourney, or other generative AI tools informally to speed up tasks, brainstorm ideas, or reframe problems. This represents what has become known as “shadow AI” or the unsanctioned use of AI tools within organizations.
This experimentation is already happening — often without formal approval. Employees use generative AI to reformulate emails, structure presentations, conduct research, and develop ideas outside any official framework. This grassroots innovation reveals something important: when official channels fail to provide useful tools, people find their own solutions.
Shadow AI reveals something important about workers: they are not passive recipients of technology but active innovators, already experimenting with AI to craft better ways of working. Yet because this happens in the shadows, it creates two distinct sets of problems.
First, there are the organizational risks that come when innovation happens outside official channels. Companies lose control over data security, can’t ensure quality standards, and face potential liability when AI-generated work contains errors. What lies in the shadows is simply hard to manage.
Second, and perhaps more troubling, are the social inequalities this creates within workforces. When only certain groups, typically those with more education, technical confidence, or workplace autonomy, feel empowered to experiment with AI, it reinforces existing workplace hierarchies. Meanwhile, other workers are left behind, unable to access the same job crafting opportunities that AI could provide.
Together, these organizational and social risks show why shadow AI, despite revealing workers’ innovation, cannot be the long-term solution. The lesson is clear: ignoring workers’ agency doesn’t make it disappear, it just makes it invisible.
AI as a constraint on autonomy
But there’s a darker side to this story especially that across different industries, we’re seeing AI systems being used to watch over workers, decide their schedules, and make their jobs more uniform. Take gig work, for example algorithms now control how much drivers earn and how their performance gets rated. When AI is used this way, workers have very little space left to make their jobs their own.
This isn’t just about keeping people engaged at work, it’s about basic rights and human well-being. When AI forces workers into rigid, unchanging routines, they lose the chance to create work that feels meaningful to them. But the damage goes deeper: constant algorithmic monitoring pushes workers into overly repetitive patterns that harm both body and mind. These systems often prevent natural breaks, create relentless performance pressure, and trap people in monotonous cycles that lead to stress, anxiety, and physical strain. Workers can no longer shape their roles around what matters to them personally. The ability to have some control over how we do our work isn’t a privilege for the lucky few, it’s a fundamental right that belongs to every worker who does their job well. When AI strips away this basic autonomy, it violates something essential about human dignity at work.
Of course, it’s not always black and white. In some situations, like hospitals or nuclear plants where safety is critical, having AI standardize certain processes actually helps everyone. The real challenge is knowing the difference between helpful standardization and going too far. Even in highly regulated jobs, workers should still be able to find ways to make parts of their work feel personal and meaningful.
The missing link: social dialogue
How do we prevent AI from stifling job crafting? The answer lies in social dialogue. This means structured conversations between employers, employees, and unions about how AI is introduced and used. Social dialogue ensures AI implementation preserves space for job crafting by giving workers voice in how technology reshapes their roles, rather than having changes imposed without consideration of how they affect workers’ ability to find meaning in their work.
Recent research from the International Labour Organisation confirms that social dialogue is most effective when it addresses three key areas:
1. employment and skill impacts of AI
2. algorithmic management practices
3. working conditions and rights in AI value chains
The study found that successful social dialogue initiatives promote what researchers call a “high road approach” to AI — one that is “based on complementing rather than replacing worker skills, empowering rather than controlling the workforce.”
Europe already offers promising examples. Spain’s Riders Law gave platform couriers the right to know how algorithms assign their tasks, following strong union advocacy. In Italy, collective agreements in food delivery have begun to regulate algorithmic decision-making. France exemplifies this approach, with its 2024 national AI Commission including union representation and specifically addressing “the role of social dialogue in supervising digital transformation at work, investments in training, evaluating algorithms, and increasing environmental transparency.”
The French DIAL-IA project demonstrates this approach in practice. The Institute of Social and Economic Research (IRES) with a co-financing of the National Agency for the Improvement of Working Conditions (ANACT) brought together around 50 participants from unions, employers, and the public sector over 18 months. On the union side, it brought together France’s major labor confederations representing workers across public and private sectors to develop resources for “technological social dialogue” at company and workplace levels, producing both a joint manifesto and practical tools launched in 2025.
These initiatives show that dialogue can turn AI from a managerial imposition into a shared resource. As Eurofound highlights, social dialogue is crucial to ensure technology adoption enhances worker well-being. Without it, AI risks reinforcing inequalities, where only skilled or privileged workers have the freedom to job craft, often through shadow AI. With meaningful dialogue, AI can be implemented in ways that expand job crafting opportunities for all workers, ensuring technology serves to enhance rather than constrain their ability to shape meaningful roles.
Policy and practice implications
Employers should involve workers early when introducing AI. Success should be measured not only in efficiency gains but also in whether AI expands opportunities for workers to engage in job crafting, actively reshaping their roles around tasks, relationships, and purposes that create meaning.
Unions and worker representatives should push for AI clauses in collective agreements, ensuring technology is designed for worker autonomy and dignity. These clauses should specifically address algorithmic transparency requirements, worker rights to understand and contest AI-driven decisions, provisions for ongoing training and skill development, and mechanisms for regular review and adjustment of AI systems based on worker feedback.
Policymakers should ensure that implementation of the EU AI Act includes worker consultation, not just technical compliance.
Regarding smaller organizations that usually lack strong union representation, alternative approaches to meaningful dialogue are essential. Some of these approaches might be employee committees focused on technology adoption, partnerships with industry associations to share best practices, or collaboration with external worker advocacy groups to ensure diverse perspectives are heard in AI implementation decisions.
The Job of Crafting a More Human AI
In order to allow AI to reshape work for the better, we must witness more than cut costs or boost efficiency by expanding the space for workers to craft meaningful roles. Shadow AI shows that workers are already trying to do this on their own, however without recognition, regulation, or fairness.
Job crafting gives individuals the brush; social dialogue provides the canvas on which everyone can paint. Only by combining the two can we ensure AI contributes to a future of work that is not just smarter but also fairer and more human.