MIT Researchers See Gradual AI Job Disruption
MIT’s FutureTech researchers studied more than 3,000 text-based tasks and had workers evaluate over 17,000 AI outputs from 40-plus models to see how automation is spreading. Their conclusion is that AI looks less like a sudden “crashing wave” of mass job destruction and more like a “rising tide” that changes tasks broadly and unevenly over time. The paper estimates AI could handle roughly 80% to 95% of text-related tasks by 2029 at minimally sufficient quality, but notes that reliable, near-perfect performance and workplace adoption still lag capability.
Hot take: this is a useful correction to apocalypse talk, but it is not a claim that AI is harmless or that job disruption is off the table.
- –The strongest finding is about timing and shape, not impact being zero: task change appears broad, gradual, and uneven.
- –The study’s “good enough” framing matters because most workplaces need reliability, not just passable output.
- –Legal and judgment-heavy work still looks harder to automate than maintenance or documentation-heavy roles.
- –The labor-market effect may lag model capability by years because integration, workflow redesign, and trust are real bottlenecks.
DISCOVERED
55d ago
2026-04-03
PUBLISHED
55d ago
2026-04-03
RELEVANCE
AUTHOR
ThereWas
