AI is replacing the jobs of the people building it
AI is displacing knowledge workers faster than physical laborers. The safest jobs right now are often the lowest paid ones.
A new analysis from Digital Planet at Tufts University’s Fletcher School maps AI displacement risk down to individual metro areas, occupations, and income levels — and the picture it produces is uncomfortable in ways that go beyond the headline numbers.
The geographic finding alone is striking. Silicon Valley, Boston, Washington, Seattle — the regions most invested in building AI — face the highest projected workforce displacement from it. San Jose leads all metro areas at 9.9% job risk. Washington D.C. tops the state rankings at 11.3%.
- The headline estimate: Around 9.3 million U.S. jobs could be displaced within two to five years, with a range of 2.7 million at the low end to 19.5 million at the high end.
- The income at stake: Between $200 billion and $1.5 trillion in annual wages, with a midpoint of roughly $757 billion.
The occupational breakdown flips the conventional automation narrative entirely. For decades the assumption was that machines replace repetitive physical work while leaving knowledge workers untouched. This data says the opposite is happening.
- Historians: 67% of tasks automatable.
- Writers and authors: 57%.
- Computer programmers and web designers: 55%.
- Roofers, orderlies, dishwashers: Under 1%.
The safest jobs right now are overwhelmingly the lowest paid ones. Not because they are protected by policy or unions, but because they involve physical, unpredictable, spatially embedded work that current AI handles poorly.
The report also identifies 33 occupations covering roughly 4.9 million workers where displacement risk could jump from under 10% to over 40% depending on how quickly AI adoption spreads. These are not gradual transitions. They are cliffs.
The report’s own uncertainty range — 2.7 million to 19.5 million — is not primarily a range of AI capability scenarios. It is a range of perception scenarios.
AI cannot do any of these jobs flawlessly. Flawless AI automation does not exist. What drives displacement is not AI reaching a reliable threshold of competence — it is decision makers believing it has. That is an entirely different and far less predictable force.
The same bias that makes casual users hand AI full autonomy and expect perfect results is operating at the boardroom level. Hiring freezes and workforce reductions are being driven by a version of AI that does not quite exist yet.
The gap between what AI is perceived to be capable of and what it actually does reliably in production is enormous. That gap is quietly responsible for a significant portion of the displacement this report is measuring.
- What the report does not include: Job creation estimates. The analysis acknowledges this omission directly, noting that whether displaced workers find new roles and on what timeline remains unanswered.
- The policy dimension: States with the highest AI exposure are legislating on the technology at four times the rate of low-exposure states. A December 2025 executive order has complicated that by directing the Justice Department to challenge certain state-level AI laws.
The Bottom Line: The perception of AI’s value is moving faster than its actual value. Markets, boardrooms, and hiring managers are making workforce decisions based on what they think AI can do. The workers paying the price for that perception gap are real, and the timeline is shorter than most forecasts have implied.
If you need on-demand GPUs for training, fine-tuning, inference, or running open-source models, give RunPod a try.
- Available hardware: H100, H200, A100, L40S, RTX 4090, RTX 5090, and 30+ more
- Cost: significantly cheaper than AWS or GCP, billed per second, no contracts
- Setup: spins up in under a minute, 30+ regions worldwide
Get the core business tech news delivered straight to your inbox. We track AI, automation, SaaS, and cybersecurity so you don't have to.
Just read what you want, and be done with it.
