29 July 2026 Heinz Süess
In the Swiss labor market, traditional entry-level jobs are disappearing, while generative AI is increasingly taking on tasks in office and knowledge-based professions. The Tages-Anzeiger reports on a gradual loss of knowledge and describes how companies are already taking countermeasures to secure young talent and decision-making capabilities within the company (Hollenstein, 2024). At the same time, a Swiss labor market analysis by Jobcloud shows that junior positions in AI-exposed functions have decreased significantly, further intensifying the discussion about deskilling (Jobcloud AG, 2026; Reuters, 2026). For HR managers and business executives of SMEs, this presents a strategic question: How can AI be used productively without undermining the development of skills within the company?
What deskilling means in the age of AI
Deskilling refers to the gradual loss of human abilities as tasks are permanently outsourced to technology. In the context of generative AI, this not only affects routine tasks but increasingly also judgment, problem-solving, and creative thinking—the very skills that are central to knowledge-intensive companies (Boston Consulting Group, 2026). In this context, the BCG report speaks of „distributed deskilling,“ where this loss of competence affects not individual individuals but entire teams or organizations (Boston Consulting Group, 2026).
This is particularly relevant for Swiss SMEs, because they rarely gain competitive advantage through pure scale effects but through expertise, customer proximity and the ability to solve complex problems pragmatically. When employees begin to uncritically adopt AI results, this often leads to more efficiency in the short term but risks compromising quality, innovation and leadership in the long term (Boston Consulting Group, 2026; Boston Consulting Group, 2024).
What the study results show
Current research does not paint a one-sided picture. Generative AI can both promote deskilling and upskilling; it is crucial to determine whether it replaces people or supports thinking. Crowston (2025) describes generative AI systems as „levellers“ because they primarily help less experienced individuals produce better results faster, thereby partially leveling traditional skill gaps. This effect can be productive, but it carries the risk of shortening learning paths and confusing apparent competence with genuine expertise (Crowston, 2025).
A frequently cited example is customer support research, which Crowston summarizes as follows: There, productivity increased by an average of 14 percent with AI support, and even significantly more so for less experienced employees; while high performers saw little benefit (Crowston, 2025). From an operational perspective, this is attractive. From a perspective of skills development, however, the question arises as to whether juniors are actually learning or merely producing better output in the short term using a system (Crowston, 2025; Generative AI and Knowledge Work, 2025).
Another study on generative AI in knowledge work shows that deskilling occurs primarily where employees are no longer forced to verify assumptions, justify decisions, or derive results themselves. Conversely, hybrid workflows, in which AI provides drafts and people consciously review, decide, and improve them, rather promote competence building than competence loss (Generative AI and Knowledge Work, 2025). Research from the education sector also reaches a similar conclusion: When AI is used substitutively, learning and skill effects decline; when it is used augmentatively, it can support the development of skills (IJIRT, 2026).
The Swiss perspective: Fewer junior roles, higher requirements
The development on the Swiss labor market is particularly concerning. According to the AI Report 2026 by Jobcloud, which is based on millions of job postings, the proportion of junior roles in AI-exposed professions has significantly decreased. Reuters reported that office and knowledge-based professions are particularly affected, and the decline has clearly been evident since the widespread introduction of generative AI (Reuters, 2026; Jobcloud AG, 2026).
This development fits with the observations from the Tages-Anzeiger article: activities such as research, documentation, initial evaluations, or draft texts are increasingly being taken over by AI today, even though these very tasks were often part of the initial career path in the past (Hollenstein, 2024). This creates a twofold problem for HR and recruiting. On the one hand, there are fewer classic entry-level roles. On the other hand, the requirements for those roles are increasing as companies are looking for candidates who, despite fewer learning opportunities, already possess more maturity, context understanding, and independence (PwC, 2026; Jobcloud AG, 2026).
PwC describes this effect in the AI Jobs Barometer for Switzerland as a shift towards more demanding profiles in AI-exposed jobs. For SMEs, this means that recruiting becomes more difficult because the talent pipeline becomes narrower, while at the same time the demand for reliable, learnable, and judgment-based employees remains high (PwC, 2026).
Why this is strategically relevant for HR and recruiting
For HR departments, deskilling is more than just a technology or efficiency issue. It concerns the question of how competencies are created within the company, how to build up the next generation of talent, and how to ensure quality is maintained in the long term. If generative AI is used extensively in recruiting, communication, or administrative processes, there is a risk that teams may work faster but lose key assessment capabilities (Boston Consulting Group, 2026; CIPD, 2025).
This concerns, for example, the wording of job advertisements, the screening of application documents, the preparation of interviews, and communication with candidates. AI can support and accelerate these tasks. However, if its results are no longer critically questioned, standardization, distortions, and a gradual erosion of professional judgment skills (CIPD, 2025; Rajaratnam, 2025) threaten to occur.
This risk is particularly high in SMEs, because HR is often poorly structured and there is strong pressure for efficiency. When few people have to handle many tasks, the temptation is great to quickly adopt AI outputs. In the short term, this saves time. In the long term, however, it can lead to a weakening of implicit knowledge, human understanding, and context assessment, even though these skills are precisely what make the difference in recruiting (Boston Consulting Group, 2026).
Five areas of action for SMEs
Establish AI policy with a focus on expertise
An AI policy should not only regulate data protection, security, and tool permissions, but explicitly address the retention of skills as well. Practice guides for HR and SMEs recommend clearly defining which tasks AI can be used for, when human review is mandatory, and in which areas AI should be deliberately avoided (Randstad, 2024; CIPD, 2025; SHRM, 2025). A comprehensible human-in-the-loop process should always be defined for HR-relevant decisions such as appraisals, sensitive communications, or final selection decisions (Boston Consulting Group, 2026; SIDD, 2026).
Design workflows as augmentatives rather than substitutives
The research evidence clearly supports using AI as an assistant rather than as a substitute for thinking. Good processes ensure that employees first understand the problem, establish criteria or formulate initial hypotheses before AI is used to generate designs, structure or variations (IJIRT, 2026; Generative AI and Knowledge Work, 2025). In recruiting, this can mean that a profile is first defined from a technical and cultural perspective and the AI is only then used to generate formulation suggestions or comparison overviews. This leaves the technical judgment to the human (Boston Consulting Group, 2024; CIPD, 2025).
Measuring and making visible skill retention
What is not visible is rarely protected in organizations. Therefore, in addition to productivity and output, companies should explicitly name the competencies that must be retained or expanded despite the use of AI. These include critical thinking, problem analysis, decision-making skills, a system understanding, and the ability to verify the plausibility of AI results (Boston Consulting Group, 2026). For HR, this may mean designing performance evaluations and learning objectives so that not only the outcome counts, but also the path to it and the quality of human decision-making (Boston Consulting Group, 2024).
Creating conscious AI-free learning spaces
Several policy and practice papers recommend creating regular situations in which employees consciously perform key tasks without AI. Such formats help actively train judgment and professional logic rather than merely assuming them (Rajaratnam, 2025; HR Dive, 2023). For HR teams, this could include manual dossier evaluations, self-drafted exit letters in sensitive cases, or the independent preparation of a structured interview. The point is not about avoiding technology, but about cognitive fitness.
Rethink youth programs
The design of learning paths for juniors, learners, and beginners in the workforce is particularly important. The Tages-Anzeiger shows, through the example of companies such as Zühlke, Thjnk, and BDO, that conscious investments in talent development are gaining in importance despite or precisely because of AI (Hollenstein, 2024). The common denominator of these examples is clear: First the foundation, then the tool. Anyone who learns accounting, text processing, analysis, or consulting only with software can often assess results less well later on. By contrast, anyone who masters the fundamentals can use AI more controlled and effective (Hollenstein, 2024; IJIRT, 2026; Generative AI and Knowledge Work, 2025).
For SMEs, a pragmatic model can be derived from this: In a first phase, basic knowledge is deliberately imparted without or with very limited AI usage. In a second phase, AI is added as an assistant. In a third phase, employees learn to systematically verify, improve, and use AI results responsibly in more complex contexts. This step-by-step model combines efficiency with competence building and is particularly practical for small and medium-sized enterprises (IJIRT, 2026; Rajaratnam, 2025).
What this means for business leaders
For business executives, deskilling is not a marginal issue in personnel development; it is a strategic leadership problem. Anyone who only looks at short-term efficiency gains risks losing the skills that enable innovation, quality, and customer trust in the medium term (Boston Consulting Group, 2026). Especially in SMEs, where knowledge is often highly tied to individual individuals and teams, collective loss of competence can have significant consequences.
Therefore, the introduction of AI should always be linked to three guiding questions:
Companies that integrate these issues early in governance, recruiting, leadership, and training not only use AI more productively, but also build their organizational resilience at the same time (Boston Consulting Group, 2024; PwC, 2026).
Conclusion
Deskilling through AI is not a natural law. It arises when companies confuse automation with the substitution of skills and prioritize efficiency over learning ability. However, the research also clearly shows that generative AI can also act as a booster for learning, quality, and development when processes are intentionally designed to be augmentative (Crowston, 2025; IJIRT, 2026; Generative AI and Knowledge Work, 2025).
For HR managers and business executives in SMEs, this represents an important leadership task. It is not the question of whether AI is used that determines success, but how it is used. Those who introduce AI today with clear guidelines, smart learning paths, and a focus on judgment, not only protect their own workforce but also strengthen the future viability of the entire company (Boston Consulting Group, 2026; PwC, 2026).
Sources
- Boston Consulting Group. (2024). Unlocking impact from GenAI: HR’s role in transforming work. BCG.
- Boston Consulting Group. (2026). When everyone uses AI, companies risk critical skills. BCG Henderson Institute.
- Chartered Institute of Personnel and Development. (2025). Use of AI in the workplace: Practical advice for HR professionals.
- Crowston, K. (2025). Deskilling and upskilling with generative AI systems. Syracuse University School of Information Studies.
- Hollenstein, E. (2024, July 27). AI in the workplace: How companies combat deskilling. Tages-Anzeiger.
- HR Dive. (2023, November 2). 7 things to include in a workplace generative AI policy.
- International Journal for Innovative Research in Technology. (2026). Deskilling generative AI higher education learning outcomes.
- Jobcloud AG. (2026). AI Report 2026: Impact of AI on the Swiss labor market.
- PwC. (2026). Switzerland AI Jobs Barometer 2026.
- Rajaratnam, R. (2025, December 9). Preventing deskilling in the age of AI: A policy agenda for competence-centred adoption.
- Randstad. (2024, July 31). How to approach your generative AI policy.
- Reuters. (2026, June 24). Fewer job offers for junior roles due to AI, Swiss study shows.
- SIDD. (2026, May 23). AI Policy Template for Swiss SMEs.
- Society for Human Resource Management. (2025, April 30). Generative AI usage policy template.
- Generative AI and knowledge work: Towards a framework for combating deskilling of knowledge workers. (2025). Master’s thesis, Halmstad University.
