For much of the modern era, higher education stood as one of the most reliable contracts a society could offer its people. One studied, one specialised, and in return one received not merely knowledge, but economic security, professional identity, and a predictable path through life. This arrangement, though never perfect, held long enough to shape entire generations. Yet in recent years, and with particular acceleration following the rise of advanced artificial intelligence systems, that contract has begun to fray in ways that are measurable, structural, and deeply unsettling.
The concern is not simply that machines are becoming more capable. That has been true for centuries. The difference now lies in the type of capability being automated. Where past technological revolutions mechanised physical labour, artificial intelligence is increasingly encroaching upon cognitive work, including the very professions once believed to be insulated by their reliance on judgement, interpretation, and human nuance. Law, psychology, and other degree-heavy disciplines now find themselves in a paradoxical position. They remain intellectually demanding fields, yet their economic foundations are being quietly undermined.
To understand the scale of this shift, one must begin with return on investment, the simple but unforgiving measure that compares the cost of education against the income it produces. In the United Kingdom, data on graduate earnings tells a nuanced story. According to the UK government's Graduate Labour Market Statistics, the median nominal salary for working-age graduates in 2024 was £42,000, but that figure spans all ages and experience levels. For those at the start of their careers, the picture is considerably more modest. Entry-level psychologists, for example, earn around £29,000 annually according to PayScale data, rising to roughly £31,000 with one to four years of experience, while the national average across all experience levels sits closer to £37,000. Law graduates on structured graduate schemes fare somewhat better at entry, with the Institute of Student Employers placing average law graduate starting salaries at approximately £43,500, though this figure is skewed heavily toward large commercial firms. Graduates entering smaller firms or non-commercial legal roles will typically earn considerably less. These figures are drawn from national salary datasets and labour market analyses. They reveal a truth that is difficult to reconcile with the cultural weight these professions carry: the financial return at entry level is modest relative to the cost and duration of the education required to reach it.
This would already be a concern in a stable labour market. However, the introduction of artificial intelligence alters the equation in a far more consequential manner. AI systems are now capable of performing tasks that were once the exclusive domain of junior professionals. In legal practice, document review, case law research, contract drafting, and elements of legal reasoning can now be assisted or partially executed by machine learning tools. Thomson Reuters, through its CoCounsel platform, has built an AI legal assistant capable of bulk document review of up to ten thousand documents, multi-step research workflows, and case timeline preparation. According to the company's 2025 Generative AI in Professional Services Report, 26% of legal organisations are now actively using generative AI, nearly double the 14% reported in 2024, with document review, legal research, and document summarisation cited as the top three use cases. These systems do not replace senior lawyers outright, but they significantly reduce the need for large cohorts of junior associates to perform the foundational work that once defined early legal careers.
This distinction is critical. The disruption does not begin at the top of the profession; it begins at the bottom. Entry-level roles, historically designed to train new graduates through repetitive but necessary work, are precisely the roles most vulnerable to automation. According to data from the National Association for Law Placement, entry-level hiring at the 250 largest US law firms decreased by 17% between 2018 and 2023, despite increased overall legal spending during that period. When that foundational layer is thinned or removed, the entire career structure above it becomes unstable. One cannot progress to senior expertise without first passing through junior experience, and if that experience is no longer economically viable for firms to provide, then the pathway itself begins to collapse.
A similar pattern can be observed within psychology, though it manifests in a subtler form. The essence of psychological practice lies in human interaction, empathy, and interpretation, qualities that machines cannot fully replicate. Yet artificial intelligence has made significant inroads into adjacent functions. Digital mental health platforms now offer automated cognitive behavioural therapy exercises, mood tracking, and conversational support at scale. Applications powered by natural language processing can provide immediate responses to users experiencing stress or anxiety, effectively handling the lower-intensity cases that might otherwise have been addressed by entry-level practitioners or support staff.
The consequence is not the elimination of psychologists, but a reconfiguration of demand. Highly specialised clinicians remain valuable, often commanding strong compensation. However, the broader base of the profession, particularly those at the beginning of their careers, faces increased competition for fewer opportunities. The floor rises in terms of required expertise, even as the ceiling remains intact. For a graduate entering the field, the investment required to reach that ceiling becomes more demanding, while the likelihood of securing stable income during the early years diminishes.
This phenomenon reflects a broader economic shift driven by artificial intelligence: the decoupling of productivity from employment. Traditionally, increased productivity within a profession led to higher demand for labour, as more work could be undertaken and more services delivered. AI disrupts this relationship by allowing the same volume of work, or greater, to be completed by fewer individuals. Research from the Federal Reserve Bank of St. Louis, based on surveys conducted in 2024, found that generative AI users saved an average of 2.2 hours per week in a standard forty-hour working week, with daily users reporting savings of four or more hours. A separate study from the London School of Economics and Protiviti, surveying nearly three thousand workers globally, found average savings of 7.5 hours per week among those using AI tools, rising to eleven hours per week for those who had received formal AI training. While time savings figures vary across studies and methodologies, the direction is consistent: AI reduces the human hours required to complete a defined scope of work.
Companies capture the majority of the value created by these efficiencies. They reduce costs, increase margins, and enhance competitiveness. Workers, on the other hand, face a different reality. As productivity per individual rises, the number of individuals required falls. This leads to wage compression, particularly in mid-tier roles, and intensifies competition for positions that remain. The result is a labour market in which being qualified is no longer sufficient; one must also demonstrate the ability to leverage AI effectively to justify one's place.
The implications for higher education are profound. Degrees were historically designed to impart knowledge and signal capability. They assumed a labour market in which human effort was the primary driver of value. In an AI-augmented economy, the value of effort is mediated by tools. A graduate who understands how to work alongside AI systems may produce significantly more output than one who does not, regardless of their academic credentials. This creates a new hierarchy, not based solely on education, but on the integration of education with technological fluency.
Institutions have been slow to adapt to this reality. Curricula often lag behind industry practices, and many programmes continue to prepare students for roles that are rapidly evolving or diminishing. The lag is not merely academic; it has financial consequences. Students invest years of their lives and substantial sums of money into degrees that may not deliver the expected return. This is not a failure of individual choice, but a systemic misalignment between education and the labour market.
It would be an exaggeration to claim that artificial intelligence is destroying the world. Such language obscures more than it reveals. A more accurate description is that AI is redistributing opportunity in a manner that is both uneven and accelerating. Those who adapt quickly, who acquire complementary skills, and who position themselves within the emerging structure may find unprecedented advantages. Those who do not may experience stagnation or decline, even if they have followed the traditional path of higher education.
The deeper concern lies not in the technology itself, but in the expectations it disrupts. Societies are built on implicit agreements, and one of the most powerful of these has been the belief that education leads to advancement. When that belief weakens, the consequences extend beyond individual careers. They affect social mobility, economic stability, and public trust in institutions. If large numbers of graduates find that their qualifications do not yield the promised outcomes, the resulting disillusionment can manifest in ways that are difficult to predict and even harder to manage.
Artificial intelligence, in this sense, acts as both a catalyst and a mirror. It accelerates existing trends, such as the oversupply of graduates in certain fields and the concentration of wealth within organisations. At the same time, it exposes the fragility of systems that relied on assumptions no longer valid. The issue is not that AI has suddenly made degrees worthless, but that it has revealed how contingent their value always was.
There remains, however, a path forward. Degrees that integrate technical competence, analytical thinking, and adaptability continue to offer strong returns. Fields that require complex problem-solving, interdisciplinary knowledge, and human judgement at a high level are less susceptible to commoditisation. Moreover, individuals who treat AI not as a threat but as a tool can amplify their capabilities in ways that were previously impossible.
Yet this path demands a shift in mindset. Education can no longer be viewed as a one-time investment with guaranteed returns. It must become a continuous process, aligned with a labour market that evolves at unprecedented speed. Institutions must adapt, individuals must recalibrate their expectations, and organisations must consider the broader implications of their pursuit of efficiency.
In conclusion, artificial intelligence is not dismantling the world outright, but it is reshaping it with a force that challenges long-standing assumptions about work, value, and education. The decline in return on investment for certain degrees is not an isolated anomaly, but part of a larger transformation. To ignore this would be to misunderstand the nature of the change. To confront it, however, requires clarity, adaptability, and a willingness to reconsider the very foundations upon which modern careers have been built.