1. Introduction

The emergence of Artificial Intelligence (AI) agents marks one of the most profound changes in the history of work. Unlike conventional automation, which focused on streamlining repetitive mechanical or computational tasks, AI agents integrate with multiple platforms and applications, execute complex workflows end-to-end, and increasingly operate with initiative and autonomy. They do not merely support human labor; they substitute, complement, and reorganize it.

The rise of these agents is rapidly altering the landscape of employment, particularly for junior professionals and white-collar workers whose activities are structured, codifiable, and therefore highly automatable. Recent research, such as Brynjolfsson, Chandar, and Chen’s (2025) study Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence, reveals that young workers in AI-exposed occupations have already experienced a relative decline of about 13% in employment compared to peers in less exposed roles. The metaphor of “canaries” aptly describes the phenomenon: early-career professionals are the first to signal the disruptive potential of AI agents in labor markets.

This article examines how AI agents are changing the world of work, explores projections for the near future, and argues for the urgent construction of a new social contract. Without proactive measures, the accelerated deployment of AI agents threatens to erode human agency, widen inequality, and undermine the very foundations of social cohesion. With appropriate governance, however, this transformation may be guided to ensure human dignity, economic security, and democratic participation.

2. The Distinctive Nature of AI Agents

AI agents differ from earlier forms of automation in three key aspects, which explain why their impact on labor structures is qualitatively different.

Autonomy

Rather than requiring step-by-step human instructions, AI agents can interpret goals, plan multi-step processes, and execute them independently. For instance, customer service agents powered by large language models (LLMs), such as ChatGPT, can now manage entire customer interactions: greeting the client, identifying the issue, searching databases, executing refunds, and escalating only when strictly necessary.

This autonomy is also manifest in commerce and payments. The Visa study “Commerce of Tomorrow, Today” envisions personal agents that detect anomalies in a car’s sensors, schedule a repair in a trusted garage, negotiate conditions according to the user’s budget, and present a simple confirmation message like “I took care of your car within your preferred schedule and budget — click here to confirm payment.” Similarly, the Itaú “IA Agente” in Brazil already integrates the Brazilian payment system called ‘’Pix’’ directly within WhatsApp, allowing customers to transfer, schedule, or automate payments without even opening a banking app.

Integration

AI agents operate across multiple platforms, combining data sources and applications into a single seamless flow. A project management agent, for example, might draft a weekly report based on Slack conversations, update a Trello board, schedule meetings in Outlook, and generate a PowerPoint presentation; all in one chain of actions.

The payments sector illustrates this potential vividly. Mastercard’s Agent Pay, developed with Microsoft and IBM, enables agents to automatically repurchase frequently used items or services based on transaction history, finalizing the entire process through a chat interface. In Brazil, the Bradesco digital bank Next, in partnership with the mobility startup Kovi, deployed the Nova agent, capable of providing real-time cost estimates for car parts and scheduling purchases online. These integrations foreshadow a near future in which “invisible payments” happen in the background, without active user intervention, as agents orchestrate financial flows with unprecedented speed and accuracy.

Learning and Adaptation

Unlike traditional systems, AI agents learn and adapt continuously. Through reinforcement learning and user feedback, they improve performance over time, often surpassing human efficiency in codifiable tasks.

For example, a legal research agent could initially provide basic case law but progressively learn a lawyer’s preferences, prioritizing certain jurisdictions, highlighting precedents in line with the firm’s strategy, and drafting arguments accordingly. In payments, Topaz’s Empática agent goes a step further, capturing users’ emotions and recommending financial actions aligned with their emotional state and consumption patterns. The Itaú Pix agent similarly adapts to customers’ routines, suggesting payment schedules and personalized recommendations.

From Tools to Superagency

These examples illustrate a common shift: AI agents no longer behave as passive tools. They are co-decision makers, interpreting context, anticipating needs, and shaping workflows.

This transition toward superagency means that decision-making capacity is now shared between humans and machines. The disruptive effect lies not only in replacing certain jobs but also in reshaping the architecture of work, commerce, and daily life. AI agents are becoming the invisible layer through which much of our economic and social activity is mediated a transformation with profound consequences for labor, regulation, and the very definition of human contribution.

3. Case Studies of Deployment

Several companies have already reported significant workforce restructuring due to AI agents:

• Klarna: Its AI assistant now handles two-thirds of customer requests, replacing roughly 700 jobs.
• British Telecom: Reports 60,000 weekly interactions led by AI agents, dramatically reducing reliance on human call center staff.
• Shopify: Downsized customer support teams after deploying agentic automation tools.

These examples illustrate that AI agents are not a distant future scenario. they are operational and reshaping labor markets today.

While displacement dominates the headlines, there is also evidence of productivity augmentation. AI-assisted call center agents resolve issues 15% faster, and creative professionals increasingly delegate research, brainstorming, and synthesis to generative agents. Workers with AI-related skills enjoy wage premiums of more than 56% (PwC, 2025).

The consequence is polarization: high-value professionals who can collaborate with AI agents thrive, while those in automatable occupations face displacement or deskilling. This bifurcation undermines the promise of shared prosperity and risks deepening social inequalities.

4. Projections for the Future

Quantitative Forecasts: Generative AI vs. AI Agents

Different institutions provide projections that vary in scale, but a clear consensus emerges: artificial intelligence, whether in the form of generative systems or autonomous agents, will profoundly reshape the global labor market. Still, it is important to distinguish between the broad impact of generative AI and the more targeted disruption brought by AI agents.

Generative AI

• Goldman Sachs (2025): Estimates that generative AI could displace up to 300 million full-time jobs worldwide over the next decade, particularly in entry-level white-collar functions such as administration, customer support, and software development.
• World Economic Forum (2025): Predicts the creation of 69 million jobs globally by 2028 but the elimination of 61 million, with net gains concentrated in AI-intensive and high-skill sectors such as data analysis, machine learning engineering, and AI governance.
• OECD (2023): Reports that nearly 27% of jobs in advanced economies are highly exposed to generative AI, but the degree of automation versus augmentation will depend heavily on institutional choices and corporate adoption strategies.
• McKinsey (2025): Suggests that generative AI could add $2.6 to $4.4 trillion annually to global GDP, but only if firms successfully redeploy displaced workers into complementary tasks.

AI Agents

AI agents represent a narrower but more acute challenge because they automate end-to-end workflows, not isolated tasks. Their disruptive power lies in their ability to connect multiple platforms, take initiative, and execute processes independently.

• Stanford Study (Brynjolfsson, Chandar & Chen, 2025): Shows that employment among 22–25-year-olds in occupations highly exposed to AI agents (e.g., customer support, entry-level programming) has already dropped by ~13% relative to less exposed groups since 2022.
• Visa & Mastercard industry pilots (2024–2025): Demonstrate how agents are reshaping financial services. For example, Mastercard’s Agent Pay automates recurring purchases and transaction approvals, while Visa’s “Commerce of Tomorrow” envisions autonomous agents managing household budgets, scheduling car maintenance, and executing payments invisibly. Such capabilities directly threaten routine finance and administrative jobs.
• PwC Global Jobs Barometer (2025): Finds that workers with AI-agent-related skills enjoy a 56% wage premium, highlighting polarization: those who adapt thrive, while others risk exclusion.
• Tony Blair Institute (2025): Projects unemployment increases ranging from 290,000 to 1.5 million in advanced economies, with AI agents accelerating short-term displacement compared to more diffuse generative AI impacts.

In short, generative AI has a broad transformative potential across industries, but AI agents concentrate their disruption on specific workflows, especially those historically entrusted to junior professionals, the traditional entry point to career progression.

Risks of Cognitive Atrophy

Beyond job loss, the rise of AI agents carries the risk of cognitive atrophy. As autonomous systems increasingly handle complex, integrative tasks (drafting legal memos, building financial models, or producing market analyses) junior workers lose opportunities to develop these competencies organically.

The Stanford research underscores this point: older, more experienced workers have remained relatively insulated so far, while younger professionals are disproportionately displaced. The risk is not only individual deskilling but also intergenerational erosion of tacit knowledge, as future cohorts may never acquire expertise that was traditionally accumulated through years of practice.

Generative AI, in turn, can exacerbate this trend by providing “shortcuts” that reduce the need for critical thinking, synthesis, and creativity. Without institutional mechanisms to ensure skill-building, the labor force risks becoming increasingly dependent on machines for core cognitive functions.

5. Psychological and Social Costs

The transformation of work also produces significant psychological and social consequences.

• EY Workforce Sentiment Report (2025): Finds that 71% of workers fear AI will replace their jobs, while 75% believe their current role may be obsolete within a decade.
• Microsoft Work Trend Index (2024): Notes that while 70% of workers would delegate tasks to AI to reduce workload, 60% simultaneously fear being replaced if they do not adapt quickly. This “dual anxiety” creates a paradox of adoption.
• Sociological surveys (WEF, 2025) highlight growing feelings of insecurity, alienation, and loss of meaning, particularly among younger professionals excluded from early career experiences.

For AI agents specifically, these fears are intensified because the technology does not merely augment; it replaces entire sequences of tasks. The example of financial “invisible payments” illustrates this shift: when agents negotiate, authorize, and reconcile transactions without user intervention, the individual’s role in economic activity diminishes, leading to a sense of dispossession of agency.

These intangible costs highlight the urgency of constructing a new social contract: one that ensures not only economic security through reskilling and social protections but also safeguards human dignity, purpose, and agency in the age of intelligent machines.

6. Towards a New Social Contract

The rapid rise of AI agents forces us to confront a fundamental dilemma: how should societies, organizations, and policymakers respond to the erosion of entry-level and white-collar jobs? These positions have historically been the gateways through which younger professionals acquired tacit knowledge, developed expertise, and built careers. If they disappear, future generations risk exclusion from the same opportunities for growth and social mobility that once structured professional life.

Rethinking the Role of Junior Professionals

Instead of treating junior professionals as expendable, we must reimagine their place in this new economy driven by AI. Their contribution can no longer be defined as low-cost execution of repetitive tasks, but rather as the cultivation of skills that machines cannot replicate. Human judgment, ethical oversight, relational capacity, creativity, and adaptability are areas where younger workers must be trained and employed. To achieve this transformation, companies and governments will need to invest in new forms of professional development that reposition junior workers as early-stage incubators of resilience, capable of thriving in close collaboration with intelligent systems.

This transition inevitably calls for a profound reform in education. Schools and universities cannot continue to rely on curricula centered on rote memorization and narrowly technical instruction. The professionals of tomorrow must be prepared to work symbiotically with machines. They will need to be fluent in the language of AI and data ethics, capable of critical and creative thinking, and equipped with emotional intelligence and teamwork skills that preserve the human dimensions of work. More than ever, education must embrace flexibility and foster continuous learning, ensuring that individuals are able to reinvent themselves multiple times over the course of their lives. Without such reforms, younger generations will be doubly penalized: displaced by automation and unprepared for the new opportunities it creates.

Corrective Mechanisms

The rise of AI agents compels us to rethink the foundations of our social and economic systems. Entry-level and white-collar jobs, historically the gateways for younger generations to acquire tacit knowledge and professional experience, are increasingly being displaced by autonomous systems. If left unaddressed, this shift threatens to sever the ladders of mobility and intergenerational continuity that have long defined the structure of work.

A new social contract must begin with a redefinition of the role of junior professionals. Rather than being treated as low-cost labor for repetitive tasks, they should be prepared to assume responsibilities that demand human judgment, ethical oversight, creativity, and empathy; qualities that machines cannot easily replicate. This requires deliberate investment in professional development, where younger workers are positioned not as substitutes for agents but as complementary supervisors, innovators, and coordinators.

Such a reorientation demands a profound educational reform. Schools and universities can no longer prioritize rote memorization or narrowly technical knowledge. Curricula must prepare graduates to collaborate with intelligent systems, focusing on AI literacy, data ethics, critical thinking, teamwork, and adaptability (McKinsey, 2025). Without such reforms, younger professionals risk being displaced not only by automation itself but also by their lack of preparation for the hybrid roles of the future. For instance, Singapore’s SkillsFuture program, which provides citizens with lifelong learning credits, offers a promising model (OECD, 2023). Similar approaches could be expanded globally to ensure continuous reskilling in an era of rapid technological disruption.

In addition to education, societies may need to consider corrective labor market measures. Quota systems have historically been used to address structural inequalities, such as gender or racial disparities. Brazil’s Lei de Cotas, for instance, requires companies to employ a minimum percentage of workers with disabilities (Brasil, 1991). If evidence confirms that junior professionals are disproportionately excluded by AI-driven displacement, similar quotas or preferential hiring pathways could guarantee them access to the market. These policies would not aim to freeze inefficiencies but to preserve opportunities for skill acquisition and intergenerational fairness.

Redistributive measures may also play a role. Experiments with universal basic income (UBI), such as the Finnish pilot that provided €560 monthly to unemployed individuals, showed improvements in well-being and reductions in anxiety, even if labor market participation remained stable (Kangas et al., 2020). In contexts of AI-induced disruption, transitional income support or wage insurance could act as buffers, preventing sudden poverty and allowing workers time to retrain.

Another path lies in regulatory interventions that ensure human participation remains central in critical domains. The European Union’s AI Act requires meaningful human oversight for high-risk AI applications, including recruitment, healthcare, and education (European Union, 2024). Extending this principle, governments could mandate human-in-the-loop approaches in finance, justice, or customer service, thereby preserving not only accountability but also professional spaces for human workers.

Incentive structures for companies are also essential. Tax credits or subsidies could be provided to firms that retain or create entry-level roles in automated industries, much like existing benefits for companies that hire minority groups or invest in green transitions (European Commission, 2022). Conversely, higher taxation of firms that replace entire segments of their workforce with AI agents, as experimented in South Korea with robot taxes, could be used to fund retraining programs (Chin, 2017).

Finally, societies must confront the deeper question of how far and how fast automation should go. The speed of adoption is not technologically determined but socially chosen. France and Germany, for example, have imposed restrictions on the deployment of fully autonomous vehicles, not only for safety reasons but also to protect jobs in the transport sector (OECD, 2021). These cases illustrate that governments can deliberately pace innovation to balance efficiency with stability. The philosophical question is unavoidable: do we wish to automate everything that can be automated, and at what cost? To rush forward without restraint risks not only unemployment and inequality but also the erosion of social cohesion and the rise of poverty and crime.

7. A Philosophical and Political Choice

Ultimately, the question of how far and how fast to automate is not purely technical. It is a philosophical and political choice. Societies must ask themselves whether everything that can be automated should be automated, and at what pace such transformations ought to occur. The promise of efficiency and cost reduction must be weighed against the risk of social dislocation, rising unemployment, poverty, and even crime. These are not decisions for engineers alone but for citizens, policymakers, and business leaders alike.

For companies, engaging with this debate is not a burden but an opportunity. Firms that pursue balanced adoption of AI agents, combining efficiency gains with investments in human potential, will enjoy more resilient workforces and stronger reputations. In doing so, they will contribute not only to their own competitiveness but also to social stability.

The new social contract must therefore be understood as both a strategic and an ethical imperative. It requires deliberate pacing of automation, large-scale reform of education, corrective mechanisms for those most vulnerable to displacement, and a shared sense of accountability across business, government, and civil society. The integration of AI agents into the workplace is not merely a technological revolution. It is a turning point in the story of human labor, a moment when we must decide whether automation will serve the many or the few. The choices made today will shape not just markets, but the dignity, security, and opportunities of generations to come.

References 

Brazil. (1991). Law No. 8.213 of July 24, 1991: Social Security Benefits Plan. Official Gazette of the Union.

Brynjolfsson, E., Chandar, B., & Chen, D. (2025). Canaries in the coal mine? Six facts about the recent employment effects of artificial intelligence. Stanford Digital Economy Lab.

Chin, J. (2017, August 7). South Korea plans to tax robots. The Wall Street Journal. https://www.wsj.com

European Commission. (2022). Guidelines on state aid for climate, environmental protection and energy 2022 (CEEAG). Official Journal of the European Union.

European Union. (2024). Artificial Intelligence Act: Regulation of the European Parliament and of the Council.

EY. (2025). Workforce sentiment report 2025. Ernst & Young Global.

Goldman Sachs. (2025). Generative AI: The productivity boom and labor market disruption. Goldman Sachs Global Investment Research.

Kangas, O., Jauhiainen, S., Simanainen, M., & Ylikännö, M. (2020). The basic income experiment 2017–2018 in Finland: Preliminary results. Ministry of Social Affairs and Health, Finland.

Klarna. (2024). Annual and sustainability report 2024. Klarna Bank AB.

Mastercard. (2025). Agent Pay: Intelligent payments for the age of AI. Mastercard Insights.

McKinsey & Company. (2025). The future of work with AI: Empowering people through superagency. McKinsey Digital.

Microsoft. (2024). Work trend index: 2024 annual report. Microsoft Corporation.

OECD. (2021). Regulatory approaches to autonomous vehicles. OECD Publishing.

OECD. (2023). Education policy outlook: SkillsFuture Singapore. OECD Publishing.

PwC. (2025). Global AI jobs barometer: Emerging trends in skills and employment. PricewaterhouseCoopers International.

Shopify. (2024). Automation and workforce restructuring update. Shopify Inc.

Tony Blair Institute for Global Change. (2025). The impact of AI on the labour market. Tony Blair Institute.

Topaz. (2024). Empática: Emotional AI for financial decision-making. Topaz Digital Solutions.

Visa. (2024). Commerce of tomorrow, today. Visa Perspectives & Institute for the Future.

World Economic Forum. (2025). The future of jobs report 2025. World Economic Forum.