Not every job will survive the artificial intelligence revolution
As generative AI systems become adept at reading, writing, and reasoning, the most exposed roles are not those on the factory floor, but in the cubicle. For decades, knowledge work promised security. Today, it faces the sharpest disruption since the arrival of the personal computer.
A white-collar turning point
The global analysis led by Evercore ISI and venture studio Visionary Future examined more than 160 million U.S. jobs and found that nearly every role carries some degree of AI exposure. But it is white-collar, information-based work that faces the greatest structural risk.
Evercore’s researchers observed that AI excels in tasks involving information ordering, memorization, and data processing—precisely the skills that anchor administrative, financial, and analytical occupations. In contrast, jobs demanding empathy, creativity, or physical dexterity remain far less affected.
Their findings align with research from OpenAI and the University of Pennsylvania, which shows that 80 per cent of U.S. workers have at least 10 per cent of their tasks exposed to large language models, while 19 per cent have more than half of their tasks potentially automatable. The OECD and International Labour Organization draw the same conclusion: clerical and routine cognitive roles face the most direct threat.
The quiet collapse of the clerical core
Clerical and administrative jobs—the connective tissue of modern organizations—are in the crosshairs. Data entry, scheduling, invoice processing, and correspondence are all activities that AI can now handle at scale.
The ILO warns that “the greatest impact of generative AI will be in clerical work.” This poses a demographic challenge as well: globally, women hold a majority of administrative positions, meaning AI’s disruption risks deepening gender inequalities unless companies move quickly to create new pathways for displaced workers.
When intellect becomes exposure
Generative AI doesn’t just perform manual calculation; it writes, summarizes, and advises. That means its influence reaches upward into higher-paying professions. Evercore ISI notes that “roles with compensation over $100,000 annually may both be impacted by, and benefit more from, AI augmentation.”
For knowledge-based occupations, the risk lies not only in job loss but in job hollowing—where AI performs the complex analysis and humans become editors or overseers. The work remains, but its creative and developmental content diminishes.
The top ten professions most at risk from AI
Based on cross-referenced findings from Evercore ISI, the ILO, OpenAI–UPenn, and OECD research, the following categories appear most vulnerable to near-term AI disruption:
- Data entry clerks
- Administrative and executive assistants
- Payroll and accounting clerks
- Customer service representatives and call-centre agents
- Paralegals and legal assistants
- Market research and data analysts
- Junior financial and investment analysts
- Technical writers and translators
- Basic computer programmers and QA testers
- Report and documentation specialists in professional services
Each of these roles relies heavily on structured information processing and standardized written output—precisely where large language models and generative tools perform strongest.
The automation paradox
Attempts to replace workers entirely with AI have so far revealed the limits of automation. As the HBR authors note, efforts to run fully automated call centres “have stumbled when confronted by novel customer issues.” Yet even partial automation changes the economics of work. If a chatbot can handle 60 per cent of customer queries, a company can scale operations with far fewer employees.
That partial displacement effect is already visible in software engineering, where tools like GitHub Copilot automate much of the code-writing process, leaving senior engineers to refine and test rather than build from scratch.
The high earners’ dilemma
While administrative roles will be hit hardest, higher-wage professions are not immune. Financial modelling, contract drafting, compliance reviews, and even aspects of strategic consulting are increasingly handled by generative systems. The IMF estimates that about 60 per cent of jobs in advanced economies will be touched by AI, with the highest exposure among educated professionals.
In practical terms, this means that the same technology boosting productivity for top performers may erode career pathways for entry- and mid-level professionals who once relied on those repetitive tasks as training grounds.
What HR must prepare for
For HR leaders, this shift demands proactive planning. First, workforce audits should identify roles with a high proportion of language-based, rule-driven tasks. Second, reskilling programs need to emphasize uniquely human abilities: interpersonal communication, creative problem-solving, and domain-specific judgment.
Evercore’s analysis suggests that on average, 32 per cent of every job function can be leveraged by generative AI tools to enhance productivity. The implication for HR is not replacement, but redesign. Tasks must be redistributed so that human workers spend less time reproducing information and more time interpreting it.
Finally, communication will be critical. Employees who feel blindsided by automation lose trust. Those who are trained to use AI responsibly—and shown how it can elevate their output—become collaborators in change.
The new frontier of job design
The authors of the HBR report argue that the question is no longer whether AI will replace jobs, but how “humans + AI” will reshape business functions. For HR professionals, that means shifting from a protection mindset to a reinvention mindset.
Roles once built around documentation or reporting will need to evolve toward judgment, context, and relationship-building. In a decade, the most valued employees will not be those who work faster than machines, but those who know when not to trust one.
AI is not the end of work—but it is the end of a certain kind of work. Clerical and mid-level professional roles, once the engine of upward mobility, are facing an identity crisis. As machines learn to think, human value will rest less on what we can process—and more on how we can decide.
