The headline numbers, and why they're less scary than they sound
Every few months a new statistic makes the rounds: AI will affect hundreds of millions of jobs. Goldman Sachs puts the global figure at around 300 million jobs exposed to AI automation in some form. The IMF estimates nearly 40% of global employment is AI-exposed, rising to roughly 60% in advanced economies. Numbers like these travel fast and land hard — but they measure something narrower than they sound like they measure.
“Exposed” doesn't mean “eliminated.” It means a meaningful share of the tasks inside a job can now be touched by AI in some way. What happens next — whether a role shrinks, transforms, grows, or stays roughly the same — depends on a distinction that matters more than the raw exposure number itself: is AI automating the task (replacing the human step entirely) or augmenting it (making a person faster or more capable at something they're still doing)?
Anthropic's own research into this question, based on real usage patterns across millions of conversations rather than theoretical capability alone, found that actual AI usage remains a fraction of what the technology is theoretically capable of — and crucially, found no systematic rise in unemployment among workers in highly AI-exposed occupations since generative AI went mainstream in late 2022. There is some evidence that hiring of younger, entry-level workers has slowed in the most exposed roles — a pattern that shows up clearly in India's own data, discussed below.
Automation vs. augmentation: the split that actually matters
Academic research tracking this divide over time (Marguerit, 2025, studying the US labor market from 2015–2022) found a consistent pattern: automation-heavy AI tends to reduce new job creation and wages in lower-skilled occupations, while augmentation-heavy AI tends to expand new work and raise wages in higher-skilled ones. In plain terms — AI that replaces a task tends to shrink the job around it; AI that assists with a task tends to grow the job around it, at least for now. That asymmetry is a real reason to think carefully about which side of the split a given role falls on, rather than treating “AI exposure” as one uniform threat.
This is also why a single “risk score” per job title, without that automation/augmentation context, can be misleading on its own — which is exactly the distinction this site's exposure checker is built to surface, rather than collapsing everything into a single number.
Who's actually affected — and it's not always who you'd expect
A striking, repeated finding across multiple 2026 studies: the occupations perceived as most vulnerable aren't always the ones actually losing ground. In the US, University of Pennsylvania and OpenAI researchers found some of the highest-earning white-collar roles were among the most exposed categories — not just routine clerical work. Brookings research found a stark gender gap in exposure: roughly 79% of employed women in the US work in occupations with high automation risk, versus 58% of men, largely because women are more concentrated in administrative, clerical, and customer-service roles — exactly where AI's task-level reach is currently strongest.
At the same time, roles widely assumed to be next on the chopping block are sometimes doing the opposite. This shows up clearly in India's IT sector, and it's worth its own section.
The India picture: less alarming than the headlines, but not simple either
India's story in 2026 doesn't fit neatly into either “AI is coming for your job” or “nothing to worry about.” Both things are true at once, for different parts of the workforce.
In the formal IT sector, the fear hasn't matched the data — so far. The most detailed study to date, from the Indian Council for Research on International Economic Relations (ICRIER) with OpenAI's support, surveyed 650 IT firms across 10 Indian cities between November 2025 and January 2026. Its central finding: generative AI is not triggering mass layoffs in Indian IT. Instead, productivity gains are outpacing productivity-driven job losses by roughly 3.5 to 1 across the business units studied. Perhaps the most counterintuitive result: roles commonly assumed to be most at risk — software developers, database administrators — are actually among the roles seeing the strongest demand growth, because AI tools have made these workers more productive rather than replacing them outright.
But entry-level hiring is where the real pressure shows up. The same ICRIER study found hiring moderation concentrated specifically at the entry level, while mid- and senior-level hiring stayed stable. Researchers were careful to note this can't be attributed to AI alone — it partly mirrors a broader post-pandemic hiring correction — but the pattern lines up with global research showing younger workers in exposed occupations face a tougher hiring market than their more experienced peers.
The skills employers want are shifting, not disappearing. 63% of Indian IT firms in the ICRIER study reported rising demand for candidates who combine domain expertise with AI or data skills — a “hybrid skillset premium,” not a demand collapse. Separately, India's tech workforce is projected to grow to 7.5 million by 2030, and AI-related job postings are expected to grow 32% year-on-year into 2026, concentrated in IT services, BFSI, and manufacturing.
Beyond the formal tech sector, the risks are more structural than technological. India's much larger informal workforce — estimated at 80-90% of all workers — faces a different kind of exposure: not necessarily direct AI displacement, but a structural vulnerability from having little training infrastructure, no formal contracts, and limited social protection to fall back on if their sector does shift. The ILO's 2026 Global Employment and Social Trends report flags Southern Asia specifically as a region where informal employment is still growing, even as the formal economy modernizes. Blue-collar sectors — manufacturing, logistics, construction, healthcare support — employ roughly 300 million Indian workers; most researchers expect AI and robotics to raise productivity in these sectors gradually rather than displace workers en masse, though this depends heavily on how quickly reskilling infrastructure catches up.
What's likely ahead
A few patterns worth watching, based on where the research consistently points:
- The gap between AI's theoretical capability and its actual usage will keep closing. Most studies agree AI is currently used well below its technical ceiling — meaning today's exposure numbers are closer to a floor than a final answer.
- Hybrid skills — domain expertise plus AI fluency — are becoming the real premium, not AI skills alone. Someone who deeply understands banking, law, healthcare, or manufacturing and knows how to work alongside AI tools is proving more valuable than either skill in isolation.
- Entry-level and routine-task-heavy roles face the most near-term hiring pressure, while experienced workers in the same fields are, for now, largely holding steady or growing in demand — a pattern that suggests reskilling support may matter most for people early in their careers.
- The informal economy — not the visible tech sector — carries the larger structural risk in India, simply because it has the least cushion (training access, savings, formal protections) to absorb any disruption, technological or otherwise.
The honest takeaway
Nobody has a reliable crystal ball here — even the researchers publishing this data are explicit that past attempts to forecast technology's labor impact have a mixed track record, and this one deserves the same humility. What the evidence does support, consistently, across US, global, and India-specific research: AI's effect on jobs today looks far more like task-level reshaping than mass displacement — at least so far, and at least in the sectors best studied. The way to navigate that isn't panic or dismissal — it's understanding, at a task level, where your own work sits on the automation-versus-augmentation spectrum, and building the hybrid skills that seem to matter most regardless of how the numbers move next. That's exactly what this tool is built to help with — start by checking your own occupation.