AI Is Not Coming for the Warehouse Worker. It Is Coming for the MBA

A new study from Anthropic shows that the workers facing the highest AI exposure today are not low-wage manual labourers — they are computer programmers, financial analysts, and customer service professionals earning nearly 47% more than the average worker.

Why This Matters

Every conversation about AI and jobs has, until now, pointed in the same direction: automation is a blue-collar problem. Factory workers, delivery drivers, warehouse staff — these are the people who should be worried. The Anthropic Economic Index report upends that assumption entirely. It shows, with real usage data from millions of Claude conversations, that generative AI is targeting cognitive and knowledge work first. If you work in an office, write code, process data, or produce analysis for a living, this study is directly about you.

What Happened

Anthropic published a labour market study built around a new concept called "observed exposure." Rather than asking what AI theoretically could do to a job, the researchers measured what it is actually doing — by analysing real Claude usage logs from enterprise deployments and mapping them to specific occupations.

The methodology combined three data sources: O*NET task data covering roughly 800 US occupations, prior task-level exposure estimates from a study by Eloundou et al., and Anthropic's own Economic Index, which tracks how Claude is being used across millions of real work conversations. The result is a coverage score for each occupation that reflects actual substitution and augmentation patterns happening inside firms right now.

The findings are stark. Computer programmers have observed AI coverage of around 75% of their core tasks. Data entry keyers sit at approximately 67%. Customer service representatives, financial analysts, and market researchers all rank in the top tier. Meanwhile, roughly 30% of workers — cooks, motorcycle mechanics, lifeguards, bartenders — show zero observed coverage. Their work involves physical presence, embodied skill, or real-world interaction that AI simply cannot mediate through software.

The study also finds that workers in the most exposed occupations earn about 47% more than those in jobs with no AI exposure, are more likely to hold graduate degrees, and are disproportionately women and Asian. This is not a story about replacing the bottom of the income ladder. It is a story about AI targeting the middle and upper rungs.

Why This Happened

Generative AI is built for text, code, data, and analysis. That is not a coincidence — it reflects where AI research effort has been directed and where the commercial demand was strongest. Enterprises were willing to pay for tools that could assist knowledge workers, and so models like Claude were trained and deployed in exactly those environments.

The Eloundou et al. framework that preceded this study estimated theoretical exposure using LLM capabilities — essentially asking whether an AI could handle a given task. That work pointed toward white-collar exposure, but it remained hypothetical. What Anthropic's study adds is the crucial step of verification: here is what is actually happening in enterprise workflows today, not what might happen in theory.

The gap between theoretical and observed exposure is significant. For "Computer and Math" jobs, around 94% of tasks are theoretically feasible for LLMs — but only about one-third are covered in current observed usage. For "Office and Administrative" roles, the theoretical ceiling is 90%, while actual coverage sits well below that. We are still in early diffusion, not full adoption. But the trajectory is clear.

What This Means

The headline employment numbers are not alarming yet. The study finds no statistically significant rise in unemployment among the most exposed workers — and the authors are careful to note they could detect a shift of even 1 percentage point if it existed. What they find instead is that overall unemployment looks similar across high- and low-exposure groups. That will give some comfort. It should give limited comfort.

Here is what the study does find that most coverage has glossed over: job-start rates for workers aged 22 to 25 entering highly exposed occupations have fallen by roughly 14% compared to 2022 levels. That decline does not appear among workers aged 26 and above. It is specific to young people entering careers for the first time.

This is the part that deserves more attention than it is getting. Firms appear to be maintaining their existing experienced staff while quietly reducing entry-level intake in roles where AI can handle junior tasks. The analyst who has been at the firm for eight years is still employed. The 23-year-old who would have been hired to do the first two years of their job — processing data, writing first drafts, running reports — may not be getting that offer anymore.

For India, the implications are specific and serious. India produces more engineering and MBA graduates per year than almost any country on earth. A significant share of the entry-level pipeline for Indian IT services firms, global capability centres, and financial services companies sits directly in the occupations showing the highest observed AI exposure — software development, data analysis, business process work, and customer-facing roles. If the pattern playing out in US hiring data replicates here, it is not experienced professionals who will feel it first. It is the 2025 and 2026 batch graduates competing for roles that AI is beginning to absorb. Campus hiring has already shown signs of stress at mid-tier IT companies. This study provides a structural explanation for why.

The deeper issue — and this is the argument I find most compelling — is that we are not just watching job displacement. We are watching career pipeline compression. Becoming a senior financial analyst in 10 years requires having been a junior analyst for the first three. If AI absorbs those junior years, the training ground disappears. The apprenticeship layer of white-collar work is not a nice-to-have. It is how expertise actually forms. Losing it does not just affect young workers today — it affects the supply of senior talent a decade from now.

What Happens Next

Watch the 2026 and 2027 campus hiring numbers at large Indian IT firms, global banks, and consulting companies. If the US pattern holds — experienced workers retained, entry-level intake reduced — you will see it show up as declining offers for fresh graduates before you see it in any unemployment statistic.

Also watch for "observed exposure" to become a standard metric. Anthropic has demonstrated that real usage data is more predictive of employment dynamics than theoretical capability estimates alone. Every 10 percentage point increase in observed exposure correlates with a 0.6 percentage point reduction in projected employment growth. That is not a large number yet — but the methodology will almost certainly be updated as AI adoption accelerates, and the signal will get louder.

The Bottom Line

The Anthropic study moves the AI-and-jobs debate from speculation to measurement. It shows that the workers most exposed right now are educated, well-paid, and in occupations that felt secure two years ago. Unemployment has not yet moved — but hiring for young entrants has. The damage is appearing at the start of careers, not the middle. For a country that trains millions of white-collar workers every year and relies on entry-level knowledge jobs as the first rung of economic mobility, that is not a distant problem. It is arriving now.

If this breakdown of AI's real impact on careers was useful, there is more like it every week.

The Gridpulse Brief lands in your inbox every Sunday morning — five stories across AI, Tech, Finance, Business, and Science, already read, already analysed, already explained. No algorithm. No noise. Just the week's most important developments and exactly what they mean for you.

Free forever. One click to unsubscribe anytime.

Subscribe to The Gridpulse Brief