How worried should I be about AI taking my job?

For most of the last decade, the dominant economic consensus was that AI would complement knowledge workers, not replace them. That consensus is eroding. Hundreds of economists and researchers — including Stanford economist Erik Brynjolfsson, a long-time techno-optimist on this question — have signed open letters and published warnings that the displacement risk is arriving faster and across more occupations than forecasters previously expected.

The shift in concern is driven by two things that changed since 2023. First, the technology: large language models and their multimodel descendants can now perform economically useful writing, analysis, coding, and decision-support tasks that were previously considered too complex for automation, and the improvement trajectory has not flattened. Second, the adoption speed: enterprises that once experimented with AI tooling are now integrating it into core workflows, and that integration is replacing entire segments of work, not just individual tasks.

The occupations most exposed right now are not blue-collar or physical-labor roles — those remain shielded by hardware costs and complex motor control — but white-collar knowledge work. Paralegal research, first-draft copywriting, junior-level software development, customer-service management, and certain financial-analysis functions are already seeing measurable substitution effects. The MIT Shaping Work initiative, which tracks AI deployment across industries, has published case studies showing that when an AI system can perform 80% of a role’s core tasks at acceptable quality, the business incentive to restructure around the remaining 20% is weak; the incentive to eliminate the role is strong.

The practical near-term advice is not to pivot into a manual trade out of fear — that’s premature — but to audit your own work for task replacement risk specifically. Which 20% of your weekly tasks could be done by a capable AI assistant today? If the answer is already above 50%, and your role’s value is tightly coupled to task completion rather than judgment, relationship, or rare domain expertise, that is a real signal. The window to build the judgment- and relationships-heavy version of your career is open but narrowing.