:max_bytes(150000):strip_icc():format(jpeg)/GettyImages-1725679453-63a72e6fe23f4d89b0ae3f32f5c21ca8.jpg)
Back in 2023, Goldman Sachs warned that generative AI could put 300 million jobs at risk worldwide. By 2025, experts warn that AI could wipe out half of all entry-level, white-collar jobs—and spike unemployment to 10%-20% over the next several years.
Large language models (LLMs) like Claude or ChatGPT can now write marketing copy, compose poetry and short stories, draft legal memos, and debug code in seconds. It can search the web, collate sources, generate research summaries, and even spit out polished slide decks. That makes many people wonder: Is my job next?
Recent research suggests the answer could depend less on your job title and more on the bundle of tasks you perform each day. Think of tasks as the sub-units of work that fill your calendar: drafting an invoice, negotiating with a supplier, sketching a storyboard frame, reconciling a ledger entry, or writing some code.
Depending on how any of these tasks can be automated with AI, you might or might not start to worry. Below, we explain how to gauge your risk and potential upside amid the AI rollout.
Key takeaways
- The more of your daily tasks that large-language models (LLMs) can already handle, the higher your displacement risk.
- Workers whose task mix ranges from easily automated to hard-to-automate will likely fare better than specialists who do one thing well.
- With the right design and policy, the technology could revive middle-skill, middle-income work rather than destroy it.
Task Exposure: The Metric to Watch
No surprise here: jobs composed mainly of tasks that AI can do entirely are most at risk. On the other hand, those that involve at least some human-only tasks appear to be safe (for now), as employees shift to the creative, client-facing, uniquely human tasks that AI still can’t do.
Run a mini-audit on yourself: list your top 10 weekly tasks and tick off any that a GPT-4-level model could do today. If AI could handle more than 50%, that signals displacement risk; under 30% suggests that AI could provide productive augmentation.
| Example Tasks-at-Risk | |
|---|---|
| Task | Likelihood an LLM Can Do It Well Today |
| Draft a marketing email announcing a new product | High |
| Translate a memo from English to Spanish | High |
| Summarize a 20-page research article into five bullet points | High |
| Proofread an article or blog post for grammar and style | High |
| Generate a first-pass legal memorandum citing precedent | Moderate |
| Build a financial model with bespoke tax rules in Excel | Moderate |
| Analyze customer sentiment from 100 call transcripts and flag hot issues | Moderate |
| Write a song or compose music | Moderate |
| Negotiate contract terms with a long-standing client over a Zoom call | Low |
| Troubleshoot a noisy car engine in the shop | Low |
| Facilitate an in-person brainstorming session for a fresh ad concept | Low |
History Says Disruption Arrives in Waves—Not Overnight
If we look to history, we find that technological disruption tends to diffuse through the labor market over a period of years. Indeed, the U.S. job market actually changed more slowly from 1990-2017 than in any earlier period, despite the arrival of computers and the internet.
For career planning, that means AI shock is unlikely to hit all at once like a meteor; instead, watch for gradual but compounding shifts. Workers who track such early indicators can pivot before the crest of the wave, much like typists who re-skilled into desktop-publishing roles during the early days of the personal computer.
We are already seeing some strong signals: sharp declines in retail jobs, stalled growth in low-paid services, rapid STEM hiring, and shrinking middle-wage employment—all of which might indicate the pace has begun to accelerate.
Where AI Augments Rather Than Replaces: A Middle-Class Reboot?
Economists argue that AI’s true promise lies in “task lifting”—the idea that software can shoulder the rote parts of complex jobs, allowing mid-skill workers to perform higher-value tasks once reserved for elite professionals.
For example, nurses using diagnostic chatbots to interpret scans or auto technicians leveraging vision models for instant fault detection.
Complementary design, however, is a choice, not a given. Researchers model three possible scenarios: no-AI, unbounded-AI with little job loss, and a “some-AI” world in which employment ultimately falls nearly 25% if firms deploy the tech purely as a labor-saving device.
The policy implications are clear: incentives such as skills-training subsidies and AI co-design grants can push firms toward augmentation scenarios that expand, rather than shrink, the middle tier of the labor market.
The Bottom Line
AI does not have to be a monolithic job killer; it can be a task-reallocation engine. Your individual vulnerability hinges on how many of your daily tasks are already “AI-ready” and whether employers deploy the technology to substitute or to complement. Audit your work, cultivate a wider task portfolio, and seek firms that invest in human-AI collaboration and you’ll be riding the AI wave rather than waiting to see whether it crashes on your career.
