Will AI take my job? A calm look at the evidence

Short answer: So far, AI has not caused mass unemployment. Economy-wide studies find no broad wave of job losses linked to AI. But the picture isn't zero either: young workers starting out in jobs that AI can easily automate, such as some customer service, software and administrative roles, have seen hiring slow noticeably. For most people, the near-term change is that their job changes, not that it disappears. The people who adapt early tend to do best.

What the big studies say

No economy-wide disruption, yet. Researchers at Yale's Budget Lab and the Brookings Institution looked at the US labor market in the first 33 months after ChatGPT launched. They found no clear link between AI exposure and changes in employment or unemployment across the economy. Jobs have been shifting a bit faster than usual, but that trend started before generative AI took off.

But early-career workers are feeling it. A Stanford study using payroll data from millions of workers found that people aged 22 to 25 in the most AI-exposed jobs saw employment fall by about 13% relative to peers in less exposed jobs, in the first version of the paper published in 2025. Later updates found the gap widening. Older, more experienced workers in the same jobs were doing fine.

The key detail: the drop was concentrated in jobs where AI automates tasks, not where it helps people do them. Where AI is a tool in skilled hands, employment held up or grew.

Why beginners are hit first

Entry-level jobs are often made of the tasks AI does well: drafting standard text, answering routine questions, writing simple code, summarizing documents. Experienced workers bring judgment, relationships and knowledge of how their organization really works, which is much harder to automate.

Companies under pressure may hire fewer juniors rather than fire seniors. That doesn't show up as layoffs, but it makes the first job harder to get.

Which jobs are most exposed

"Exposed" means AI can do a large share of the tasks, not that the job will vanish. Roles with high exposure include:

• customer support and call centers
• junior software development and testing
• translation and transcription
• data entry and some bookkeeping
• marketing copy and content production
• paralegal and document review work

Jobs with lower exposure involve physical work, in-person care or complex human interaction: nursing, skilled trades, teaching young children, social work, hospitality, many kinds of repair.

"Will AI take my job as a lawyer / accountant / engineer?"

These are among the most searched versions of the question. The short version for each:

• Lawyers: AI is taking over research, first drafts and document review. Judgment, negotiation, court work and client trust remain human. The risk is mostly to junior and support roles.
• Accountants: routine bookkeeping and reconciliation are increasingly automated. Advisory, audit judgment and tax strategy are not.
• Software engineers: AI writes a growing share of code. Demand is shifting from writing code to designing systems, reviewing and owning results. Entry-level hiring is where the squeeze shows most.
• Sales: AI handles prospect research and email drafts. Relationships and closing are still human.

What you can do now

• Learn to use AI well in your own field. The worker who uses AI to do the job better is far harder to replace than the job itself.
• Move toward judgment. Volunteer for work that involves decisions, clients, exceptions and responsibility.
• Build relationships. Inside and outside your organization. They are not automatable.
• Document your impact. Show results, not just tasks completed.
• Keep learning in small steps. A short course or a weekend project with a new tool is enough to stay current.
• If you're early in your career, look for roles with mentoring and real responsibility, and build a portfolio that shows what you can do with AI, not instead of it.

A note on "AI job calculators"

Websites that estimate the chance AI will take your job usually match your job title against lists of tasks. They can be a fun starting point but they can't see what you actually do every day. Two people with the same title can have very different exposure.

The bottom line

The data so far says: no robot apocalypse, but a real squeeze at the bottom of the career ladder in some fields. Researchers stress these are early snapshots, and things can change fast. The best protection is not to compete with AI on the tasks it does well, but to become the person who uses it to do more.

Based on published research as of October 2026.