AI & the Job Market: Revolution or Crisis?

Lisa Ernst · 27.11.2025 · Technology · 8 min

The impact of artificial intelligence on the job market is the subject of intense research. A recent MIT study, part of "Project Iceberg," sheds light on the share of the US wage bill that could technically be replaced by today's AI systems. The results show that around 11.7 percent of the wage bill in the US, particularly in office, finance, and tech professions, could theoretically be taken over by AI. However, this figure is a technical indicator, not a forecast of immediate job losses.

Source: Own Representation

AI and the Job Market: An MIT Study

A new study by the Massachusetts Institute of Technology (MIT), , in collaboration with the Oak Ridge National Laboratory (ORNL), is building a digital twin of the US labor market. This includes 151 million employees in 923 professions, distributed across around 3,000 counties, modeled with over 32,000 individual skills ( MIT Study). ). In parallel, researchers are mapping more than 13,000 software, AI, and automation tools, including generative AI, and assigning them the skills that these systems technically master ( MIT Study).

). From this combination, the "Iceberg Index" emerges, which measures the share of the wage bill in a profession that could theoretically be taken over by existing AI systems. This is the technical exposure, not the actual adoption by companies or a timeframe for implementation ( MIT Study, arXiv).

). In contrast, the "Surface Index" only reflects the AI usage observed today in tech professions and related roles and amounts to 2.2% of the wage bill, about $211 billion ( MIT Study). ). Media outlets like NDTV summarize this 11.7% concisely as "AI can replace 12% of US workers." However, the study itself emphasizes that it concerns technical capabilities, not scheduled layoffs.

The 11.7% indicates that AI systems exist today for tasks in this scope that can perform these activities at a comparable quality level ( MIT Study). ). Whether these tasks are actually automated depends on three hurdles: technology must be reliable enough, legally and organizationally justifiable, and above all, economically sensible ( MIT CSAIL, MIT IDE).

An earlier MIT study on computer vision applications found that only about 23% of labor costs in the "AI-exposed" visual tasks considered could actually be automated cost-effectively. The rest fails due to acquisition, integration, and maintenance costs ( MIT IDE, Euronews). ). This means that in many cases, it is more expensive or riskier to replace people than to supplement their work with AI. Forbes summarized this: "AI will not steal your job anytime soon."

The MIT study clearly shows that the Iceberg exposure is not highest where AI is most talked about in the media, but in classic "white-collar jobs" ( MIT Study). ). According to the Iceberg Index, the following are particularly affected:

• Administrative and support roles, such as case management, HR administration, or simple HR screenings, where many standardized documents are reviewed, texts are generated, or data is matched ( MIT Study, NDTV).
• Finance and controlling tasks, where reports, forecasts, and analyses are highly data- and text-based, for example in credit scoring, risk reports, or portfolio reporting ( MIT Study.
• Professions in professional services, such as legal research, consulting analyses, and parts of software development that rely on recurring patterns ( MIT Study, WEF).

World Bank and OECD analyses confirm that routine information work activities such as data entry, simple bookkeeping, standard customer service, or certain back-office functions are among the most automatable roles worldwide ( OECD).

Source: schulz-beratung.de

Global Forecasts: The Impact of AI on the Job Market in Different Regions.

Comparison with Other Analyses

The current MIT analysis is part of a wave of studies quantifying the employment effects of AI, but it comes with an important shift in perspective. The ILO estimates that about a quarter of jobs worldwide could be affected by generative AI to a relevant extent, with a focus on transforming tasks rather than their complete elimination.

The " Future of Jobs Report 2023" " by the World Economic Forum estimates that globally, 69 million jobs will be created and 83 million eliminated in the next five years – a net loss of 14 million jobs, with a total of 23% of today's tasks undergoing significant change ( WEF).

McKinsey estimates the potential of generative AI to contribute an additional $2.6 to $4.4 trillion annually to global GDP, if companies widely adopt AI and the released time is redirected to other value-generating activities. At the same time, the same research group notes that by 2030, up to 30% of hours worked today could be automatable if companies consistently restructure processes ( McKinsey).

The MIT Iceberg study differs in two points:

• Firstly, it shifts the focus away from broad job categories to a finely granulated skill model that links 32,000 individual skills and 13,000 AI tools ( MIT Study).
• ). Secondly, it deliberately separates technical exposure (11.7%) and current adoption (2.2%), thus providing a tool with which policymakers can simulate how different regulatory, educational, and investment strategies would influence development ( MIT Study, Chosun).

). While some media outlets draw the alarmist conclusion "12% of jobs replaceable today," the core message is rather an invitation to data-driven policymaking: Where are the greatest risks for regions, professional groups, and age cohorts, and which retraining and further training paths are realistic? NDTV, CNBC and others pick up on this figure, but themselves refer to the Iceberg Platform as the technical basis.

Impact on Employees

For employees, the Iceberg result means that in many knowledge-intensive professions, an increasing portion of routine tasks will be supported or taken over by AI systems. McKinsey-Umfragen already show that generative AI significantly accelerates writing tasks, research, and standard analyses.

International comparative data from the OECD shows that employees who use AI in their work are more satisfied on average, but at the same time may be exposed to stricter controls, higher work pace, and new data protection risks.

ILO explicitly warns that the negative labor market effects of generative AI could disproportionately affect young people, especially when entering highly digitized professions such as programming, online marketing, or digital assistant roles.

The question shifts from "Will I be replaced?" to "Which parts of my work are standardizable – and where can I consciously build skills that are difficult to automate?" These include complex coordination with clients, creative problem-solving, negotiation skills, interdisciplinary coordination, but also the competent management of AI systems themselves – prompt design, quality control, handling data protection and fairness. Companies are already reporting the emergence of new roles related to AI governance, data curation, and "AI Operations" ( McKinsey).

Recommendations for Action

For companies, the Iceberg study is not a call to quickly reduce staff, but an invitation to view their value creation in a task-oriented manner. The OECD shows that companies that successfully implement AI typically combine three things: They clearly define which tasks should be supported by AI, they invest in further training, and they adapt processes and responsibilities.

Especially small and medium-sized enterprises, according to a recent OECD-Studie , can achieve more with fewer resources through generative AI – provided they close skill gaps within the team and implement appropriate protective measures for data and employees.

In practice, this means:

• Map tasks and processes, not just job profiles.
• Start pilot projects where there is clear added value and a good data situation – for example, in reporting, customer communication, or internal knowledge databases.
• Involve employees early on to jointly define AI usage, quality controls, and limitations.

The Iceberg Platform offers an example of how exposure can be visualized at the skill level – even if it is currently focused on the US.

Source: kettner-edelmetalle.de

AI at the Conference Table: Strategic Integration and Decision-Making in Companies.

The new MIT study will further fuel debates about regulation, retraining, and social security because it provides concrete numbers for the abstract slogan "AI endangers jobs." International organizations have long been calling for active management of this transformation. The ILO recommends a combination of active labor market policies, targeted training programs, and strengthening employee rights regarding algorithmic decision-making and monitoring systems.

The OECD Employment Outlook 2023 emphasizes that mass unemployment due to AI, given historical experience, is not the most likely main danger – but inequalities are, if qualification offerings, co-determination, and social security do not keep pace.

At the same time, current reports, for example on the UK-Arbeitsmarkt, , show that without countermeasures, millions of primarily low-skilled jobs could be lost due to automation, while new, higher-skilled jobs emerge, to which many of those affected today have no direct access.

Against this background, the Iceberg Index provides a tool that allows funding programs, retraining, and regional structural policies to be more precisely aligned with those occupational groups where AI exposure and social vulnerability coincide ( MIT Study).

Conclusion

The headline "AI can replace 12% of jobs" sounds dramatic but is too simplistic. The MIT study primarily shows how large the theoretical overlap between today's AI capabilities and human activities already is – and how little of it has visibly arrived in the labor market so far ( MIT Study, NDTV).

). Whether this 11.7% translates into actual job losses, new hybrid roles, or productivity leaps depends not on an inevitable technological trend but on decisions: in companies, in politics, and in personal further education.

Anyone who wants to seriously think about the impact of artificial intelligence on the job market should therefore not read the new MIT analysis as a finished forecast, but as a high-resolution map: It shows where the front lines between automation, upskilling, and new work are shifting – and where it is worthwhile to invest in people today instead of just managing the costs of poorly controlled transformation tomorrow.