Since artificial intelligence left the laboratories and entered organizations, there have been some rather schizophrenic discussions about its impact on employment, ranging from fascination with the technology to fears of massive job losses.
In contrast to this sometimes anxiety-provoking discourse, the latest 2025 Global AI Jobs Barometer from PWC is rather reassuring.
According to the report, AI, far from destroying jobs, will create more, make them more productive, better paid, and more skilled.
But how much is this observation really worth? Is this a strong trend or, once again, a marketing review perfectly calibrated to sell transformation support services? While the study is full of interesting and sometimes impressive figures, it carefully avoids certain topics that, in my opinion, deserve some attention. Because the real issue here, in my view, is as much about the transformation of employment as it is about how we think about the governance of work in the age of AI agents.
In short:
- The PWC study asserts that AI generates economic and wage growth in the most exposed sectors without causing massive job losses, but this dynamic remains limited to certain areas and does not reflect the economy as a whole.
- AI is transforming jobs by shifting complexity rather than replacing workers, but this transformation requires significant efforts in training, organization, and governance, which are often underestimated or invisible.
- Job creation in sectors exposed to AI is slower than in others, which the study presents as an asset in a context of demographic decline, but this raises questions about the real meaning of this development.
- The rise of skills at the expense of qualifications is presented as democratization, but it poses real challenges for HR functions in terms of skills management, anticipation, and governance.
- The vision promoted by the report is based on conditions that are rarely found in today’s organizations and largely ignores the human effects of AI on work, such as stress, isolation, and feelings of injustice.
AI boosts productivity but not for everyone
The barometer’s key message is that sectors most exposed to AI (software, finance, professional services) are seeing three times higher growth in revenue per employee than those least exposed. Salaries are also growing twice as fast, even in the most automatable jobs. Added to this is the fact that all sectors, including construction and agriculture, are increasing their use of AI.
In other words, AI is not destroying jobs on a massive scale, but transforming them. That said, when we look more closely, this growth dynamic remains concentrated, both geographically and sectorally. The study therefore measures potential exposure to AI rather than its actual adoption, which is known to be much more nuanced. It is therefore not an objective reflection of the economy as a whole, but rather a snapshot of the most digitizable sectors.
Increase or displacement of work?
The report features two personas: Amina, an analyst augmented by AI agents, and John, a customer support agent whose simple tasks have been automated. In both cases, AI frees up time to focus on more complex activities. The message is clear: AI does not replace, it augments rather than transforms (AI in the workplace: going beyond augmentation to actually transform).
But this fairly consensual storytelling ignores a reality that all businesses that have gone down this route have faced: this augmentation work does not happen by itself. It requires training, reorganization, coordination, and explanation. What’s more, the skills required are evolving 66% faster in jobs exposed to AI than in other jobs, and it is the most automatable jobs that are undergoing the most significant upheaval.
The question is therefore not so much whether AI is eliminating jobs, but rather where the complexity of work is shifting. Supervising an AI agent, rephrasing a prompt, adjusting an automated decision: all of this is work, often invisible, but its scale is greatly underestimated. However, this part of the transformation usually takes place implicitly, without a clear framework or recognition.
Job creation slowing but socially acceptable
The study aims to be reassuring: the number of jobs is increasing in almost all occupations exposed to AI. However, it also points out that this growth is significantly lower than in occupations not exposed to AI and attempts to explain this with a demographic argument: in a world where the working population is declining, sluggish growth in jobs exposed to AI is actually good news in terms of avoiding tensions in the labor market.
This is where the reasoning is open to debate, as we are no longer talking about opportunity but lesser evil. Suddenly, AI is no longer seen as a lever for prosperity but as a palliative for demographic aging. This is worth noting because we are witnessing a paradigm shift: fewer jobs, but potentially better paid, better equipped, and more skilled. All this, of course, provided you are on the right side of the transition.
The end of degrees and the triumph of skills
Another point rarely mentioned in studies and the media is that in jobs exposed to AI, the demand for degrees is declining faster than in other sectors. What matters now is the ability to learn, adapt, collaborate with agents, and think in terms of systems. The study sees this as a democratization of expertise, but fails to ask who will support this transition.
For HR functions, however, the stakes are high. Managing real-time skills development plans while transformation cycles are accelerating (the lifespan of a key skill is now estimated at 12 to 18 months) is a major challenge. The triptych “Buy, Build or Bot” (hire, upskill or rely on technology) seems relevant on paper, but remains illusory without skills governance and dynamic role mapping.
An appealing but unrealistic vision?
The report’s big promise is the advent of a digital workforce orchestrated by cooperating AI agents capable of planning, executing, learning, and interacting. In theory, this sounds like a dream made possible by algorithms, but in practice, it requires:
- high-quality data,
- interoperable platforms,
- open architecture,
- seamless business/IT coordination,
- and clear governance.
The report is already unclear about who does the orchestrating. AI, humans, or a mix of both? There is a similar sense of vagueness about how this will happen and what governance is needed. It’s a bit like the confusion at Moderna (HR/IT and the reality of working at Moderna: the unspoken truths of a reorganization).
But to come back to these prerequisites, we need to be clear-headed enough to recognize that very few organizations are currently in a position to meet them. The transformation to large-scale agentic AI is much more organizational than technological. Without a solid architecture and a cross-functional vision, businesses risk multiplying isolated use cases without ever transforming the value chain.
Worse still, we currently have no way of measuring the impact of the potential asymmetry between efforts to treat AI as if it were a full-fledged employee and the fact that, in return, human employees are being neglected (IT is becoming the HR department for machines, but who is taking care of the humans?).
The gray area of things that the barometer does not measure
If I had to criticize one other thing about the report, it would be its silence on the impact on employees’ lives. There is no mention of the potential cognitive load induced by AI, nor of the side effects: dilution of responsibilities, blurring of the human role, increased demands without the means to meet them.
Far from improving the lives of all workers, AI can also make them feel more isolated, more exposed, and more stressed.
The report talks about value but not about a sense of justice or, rather, injustice. It talks about income but not about recognition, about skills but not about working conditions, and it highlights productivity gains but without saying for whom or at what cost.
It is nevertheless surprising to produce a barometer on jobs while sidestepping the human dimension of the transformation to such an extent, unless this is an implicit message.
Bottom line
PWC concludes by calling for the future of work with AI to be “intentionally designed.” We can only agree with this point, because the question is not whether we will adopt AI, but rather for what purposes, with what safeguards, in what kind of work organization, and in the service of what vision of progress.
Faced with a technology that is as powerful as it is fast, the choices to be made are political, not technical. It is not enough to equip employees; we must also equip the collective, restore meaning to work, and build the conditions for trust (Employees Won’t Trust AI If They Don’t Trust Their Leader).
Without this, AI risks being just another pipe dream, in a long series of promised but never truly achieved transformations.
Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI)
Related articles:
Will AI replace juniors? The false debate that’s only the tip of the iceberg
How to reinvent work and professions in the age of AI
AI and talent management: a dynamic approach that promotes internal mobility
Will AI change the way performance is measured and rewarded?
HR and IT merger: Moderna redesigns its organization for and with AI
IT is becoming the HR department for machines, but who is taking care of the humans?
