Basic tasks, which used to be assigned to junior staff – allowing them to learn on the job – are gradually being carried out by AI, and companies are cutting costs… and also undermining their own future, as junior staff will have fewer opportunities to progress to senior roles. (Photo: Shutterstock)

Basic tasks, which used to be assigned to junior staff – allowing them to learn on the job – are gradually being carried out by AI, and companies are cutting costs… and also undermining their own future, as junior staff will have fewer opportunities to progress to senior roles. (Photo: Shutterstock)

Tech giants are promising to train 95 million people in artificial intelligence. Yet their own report paints a picture of a market that favours specialists and demands ever more experience. As entry-level roles become automated, the path to becoming a senior professional is narrowing. Junior staff looking for their first job… will need to retrain, say these ten tech giants.

That first job wasn’t just a lower-paid role. It was where you learnt to become a senior employee. By automating some of the basic tasks whilst simultaneously raising the bar for those remaining, tech companies may well be removing the first rung from their own ladder.

This finding appears in a report published by the AI Workforce Consortium, led by Cisco in partnership with Accenture, Cornerstone, Eightfold AI, Google, IBM, Indeed, Intel, Microsoft and SAP. These ten companies analysed 50 job roles based on job advertisements published in G7 countries between July 2024 and June 2025. They aim to prepare employees for the integration of artificial intelligence (AI) and have set a collective target of training or reskilling 95 million people over ten years.

Their study highlights, above all, the spread of AI skills. These appear in 78% of the roles analysed. Seven of the ten technology roles experiencing the strongest growth are directly linked to them. Job vacancies for AI risk and governance specialists have risen by 234% in a year, those for natural language processing engineers by 186%, and those for AI and machine learning engineers by 145%.

Fewer than one in five job vacancies requires “no experience”

However, these percentages do not reflect a general boom in technology jobs. The report does not provide either the absolute number of vacancies behind each increase or their full breakdown across the seven countries. A role that was initially very rare may show triple-digit growth without actually representing a large number of recruitments.

The sample itself is not representative of the labour market as a whole. The consortium members selected the 50 occupations based on their importance to their own AI strategy and the priority they give them in their training programmes. Forty of these are ICT roles. The remaining ten cover areas such as compliance, law, human resources, finance, marketing and training. Administrative, creative and editorial professions most exposed to AI remain largely outside the scope of the study.

Even the figure of 78% needs to be refined. Of the 50 occupations, 24 are still only at the “initial” stage of integration: AI skills feature in more than 10% of job vacancies, but never exceed 25%. Seven roles have reached a “significant” level of integration, ranging from 25% to 50%. None fall within the 50% to 70% range. The eight roles where AI exceeds 70% are mainly those whose job titles are already specialised in this field. The figure therefore measures the prevalence of a new requirement in job adverts, not the complete transformation of 78% of jobs.

The most revealing part of the report lies elsewhere. The proportion of technology roles requiring at least five years’ experience rose from 37% in the second quarter of 2022 to 42% in the second quarter of 2025. At the same time, the proportion targeting candidates with two to four years’ experience fell from 46% to 40%.

The entry barrier remains high. Of the US technology job vacancies that specified an experience requirement in the second quarter of 2025, only 18% were open to candidates with one year’s experience or less. Three years earlier, this figure stood at 17%. The relative share of positions explicitly aimed at entry-level candidates has therefore not plummeted, but remains at a low level within a market that has contracted significantly.

The fallout from the pandemic and rising costs

The study, published separately by Indeed Hiring Lab, shows that this trend towards stricter requirements is specific to the tech sector. In other US professions, the proportion of job adverts requiring at least five years’ experience fell from 16% to 11% between the second quarters of 2022 and 2025. Tech employers are bucking the trend: they are seeking more experience, whilst other sectors are demanding less.

AI is a possible explanation, not a proven cause. Indeed points out that tech job vacancies had already begun to plummet by mid-2022, before ChatGPT was made publicly available. The fallout from recruitment during the pandemic, rising interest rates and the economic slowdown also played a part. Nearly half of the decline observed since the market peak had already taken place before the arrival of generative AI.

But the timing is not coincidental. Experience requirements began to tighten in early 2023, following the launch of ChatGPT, whilst this trend was not evident in the rest of the economy. Indeed puts forward two hypotheses: tech employers may be automating certain tasks previously entrusted to early-career employees, or offering higher pay to experienced professionals as they integrate AI into their operations. The data does not yet allow us to distinguish between these two mechanisms.

A sluggish market

However, they show that growth in AI roles is taking place against the backdrop of a sluggish market. As at 11 July 2025, US technology job vacancies were 36% below their February 2020 level, according to Indeed Hiring Lab. Of the 149 technology job titles with at least 1,000 vacancies at the start of 2025, only 28 had exceeded their pre-pandemic levels. Vacancies for software engineers, which remain the most numerous, had fallen by 49%.

Roles directly related to AI are among the exceptions. Job vacancies for machine learning engineers were still 59% above their early 2020 level, despite a 47% fall from their 2022 peak. Growth is, therefore, not spread across the entire sector. It is concentrated amongst a group of specialists capable of developing, integrating, securing or governing AI systems.

The latest data reinforces this trend towards senior roles. In May 2026, the number of senior-level job vacancies in the US had risen by 14.7% year-on-year, whilst vacancies for entry-level roles had fallen by 7.5%, according to a new analysis by Indeed. In software development, 69.3% of vacancies were classified as senior roles in the first quarter of 2026. Entry-level roles accounted for just 4.5%. This new study focuses on the United States and uses a methodology based on responsibilities and required training, rather than solely on job titles or the number of years’ experience required. It cannot, therefore, be directly applied to Luxembourg.

Training… already a requirement for juniors

The question raised, however, goes beyond the number of available posts. The tasks assigned to junior staff were part of their learning process: writing a first draft of the code, documenting it, testing it, handling a simple support ticket, cleaning up data, carrying out research or preparing an analysis. These tasks gave them access to production systems, errors, colleagues’ corrections and the trade-offs that cannot be learnt in a classroom.

Automating some of these tasks can improve immediate productivity. An experienced professional equipped with AI tools can produce more and oversee several operations. However, this approach creates a long-term contradiction: organisations need senior staff capable of verifying results, understanding the context and taking responsibility for decisions, whilst potentially reducing the opportunities for junior staff to gain this experience.

The consortium’s response is through training. It advises young people to master AI tools, obtain certifications, build a portfolio, carry out projects and seek out work placements and mentors. This advice shifts part of the burden onto the candidate. A certification demonstrates knowledge of a tool. It is no substitute for the experience of dealing with a security incident, deploying software to clients or working on a project that fails.

What about the Grand Duchy?

The figure of 95 million people to be trained should be interpreted with the same degree of caution. It adds together the commitments announced by several members, but the report does not publish any standardised rates for completion, certification, job placement or pay progression. Nor does it specify how many beneficiaries were subsequently recruited by the consortium’s member companies. Training and recruitment remain two distinct processes.

The report does not contain any data for Luxembourg. It does, however, put forward a hypothesis to be tested amongst banks, insurance companies, financial services firms, the Big Four, technology companies and public sector organisations: are they recruiting fewer junior developers, analysts or technicians, whilst seeking more experienced candidates in cloud computing, cyber security and AI governance? Changes in Adem job vacancies, work placements and graduate placement figures would help to gauge whether the same phenomenon is affecting the Grand Duchy.

Companies can immediately reduce their need for entry-level staff. However, they cannot permanently eliminate the stage through which these entry-level staff become the experts they claim to be lacking. The skills shortage predicted by the tech giants could also be the result of a market that no longer allows enough time to develop those skills.