Which Sectors Are Most at Risk from AI? Key Insights from Anthropic’s Labour Market Study

Overview

A new Anthropic study, “Labor market impacts of AI: A new measure and early evidence”, analyses how large language models (LLMs) like Claude are affecting jobs. It introduces a metric called “observed exposure”, which combines what AI can theoretically do with how much it is actually being used in real workplaces.

The findings show a clear divide: digitised, knowledge-based roles face higher AI disruption risk, while manual and physical jobs are relatively insulated. The patterns, though based on US data, have strong implications for India, especially for IT services and BPO.

What Is “Observed Exposure” to AI?

Anthropic’s framework distinguishes between:

• Theoretical AI coverage: Tasks that AI could perform based on its capabilities.
• Observed AI coverage: Tasks that AI is actually performing in professional use.

For example, in computer and math occupations:

• LLMs are theoretically capable of handling about 94% of tasks.
• In practice, Claude currently covers only about 33% of those tasks in observed usage.

This gap shows that while AI has immense potential, real-world adoption is still partial, especially in roles requiring physical presence or specialised practical skills.

Sectors Most at Risk from AI

Jobs that are highly digitised, routine, or information-heavy are the most exposed. According to the Anthropic study, sectors with high theoretical and observed AI coverage include:

• Business & Finance
• Management
• Computer Science & IT
• Math
• Engineering
• Legal
• Office & Administrative roles
• Sales (especially digital and CRM-heavy roles)
• Customer service

These occupations are projected to see slower employment growth over the next decade because a large share of their tasks can be automated or assisted by AI tools.

Who Is Most Affected Within These Sectors?

• Women: AI exposure disproportionately affects female workers.
  • Women hold about 54.4% of high-exposure jobs, versus 38.8% in unexposed roles.
• Highly educated professionals:
  • Graduates are nearly four times more represented in high-exposure occupations.
• Younger workers (22–25 years):
  • Hiring for this age group in high-exposure sectors has already declined by around 14%.

Even though these workers are generally better paid and more educated, they face significant adjustment risks as AI tools increasingly take over routine cognitive tasks.

Sectors More Insulated from AI Disruption

On the other side, AI’s theoretical usefulness is lower in jobs that require physical labour, hands-on skills, or direct interpersonal and care-based interaction. The study highlights sectors that are relatively less exposed for now:

• Construction
• Agriculture
• Protective services (e.g., police, security)
• Personal care (e.g., caregivers, beauticians)
• Installation & Repair
• Grounds maintenance
• Food & Serving
• Transportation
• Healthcare support (many hands-on roles)

These occupations involve manual work, on-site presence, or complex human interaction, making end-to-end automation difficult, at least in the near term.

Key Inference: Digitised vs Manual Work

The central message of the Anthropic report is:

• Jobs involving digitised workflows (documents, data, code, online systems) are at greater AI risk.
• Jobs dominated by manual or physical labour (e.g., plumbing, construction, many farm tasks) are relatively less exposed to AI disruption in the short to medium term.

Implications for India

Although the study is based on US labour market data, the patterns are highly relevant for India, where AI is rapidly spreading across sectors.

Agriculture: Low Disruption, High Augmentation Potential

In Indian agriculture, AI is more likely to augment rather than replace farmers:

• Using satellite imagery, weather forecasts, soil data, and crop patterns, AI can help farmers:
  • Decide what to sow, when to sow, and how much input to use.
  • Optimise irrigation and fertiliser use.
  • Get early warnings about pests and diseases.
  • Reduce risk and improve yield through more precise, data-driven farming.

The Union Budget 2026–27 proposed Bharat-VISTAAR (Virtually Integrated System to Access Agricultural Resources):

• A multilingual AI tool that will:
  • Integrate AgriStack portals with AI systems.
  • Connect with Indian Council of Agricultural Research (ICAR) packages of practices.
  • Offer customised advisory support to increase farm output and cut risk.

Here, AI acts mainly as a decision-support system, not a direct substitute for human labour.

IT Services and BPO: High Risk of Automation

In contrast, India’s formal, digital-first sectors face serious disruption:

• A NITI Aayog report, “Roadmap for Job Creation in the AI Economy”, estimates that:
  • Over 60% of formal sector jobs in India, especially in IT services and BPO, are susceptible to automation by 2030.
  • These sectors currently employ over 6 million people.
• The workforce challenge is compounded by:
  • Limited mathematical and scientific skills in large segments of the population.
  • Relatively low spending on education, research, and development compared with the US and China.

Market Signals: Pressure on Indian IT Stocks

Investor behaviour reflects these AI-related concerns:

• In the past year, the Nifty IT index and major stocks like TCS, Wipro, and Infosys have fallen by nearly 20%.
• Other leading IT services firms have also seen sharp sell-offs.
• A key trigger was Anthropic’s February launch of a suite of workplace automation tools capable of performing many tasks once handled by humans or traditional software platforms.

These developments highlight how quickly enterprise AI adoption can reshape expectations around future demand for IT and BPO services.

Conclusion

Anthropic’s study shows that AI’s labour market impact will be uneven:

• Most at risk: Digital, white-collar roles in finance, management, IT, legal, office administration, and customer service.
• Relatively protected (for now): Physical, manual, and care-based occupations like construction, agriculture, protective services, and personal care.
• Disproportionate effect: High-exposure roles are more likely to be held by women, highly educated professionals, and younger workers.

For India, this means preparing its workforce—especially in IT and BPO—for rapid reskilling, upskilling, and adaptation, while using AI as a tool to augment productivity in sectors like agriculture rather than replace human labour.

Source: Indian Express