If working from home feels like a long-term comfort, a senior voice from the AI world suggests it may not stay that way forever. Shane Legg, co-founder and chief AGI scientist at Google DeepMind, has said rapid advances in artificial intelligence could quietly undercut many remote jobs that exist entirely online.
Legg argued that roles based purely on cognitive work — thinking, analysing, writing or coding through a computer — are likely to face the earliest disruption.
Legg’s concern is not about offices versus homes, but about how work is done. Jobs that do not require physical presence and rely only on mental effort are easier for AI systems to replicate or assist at scale.
As AI tools become more capable, companies may find they can deliver the same output with far fewer people. In software development, for instance, Legg suggested teams of around 100 engineers could eventually shrink to about 20, with AI taking over large parts of routine cognitive work. If that happens, remote roles and junior positions could be among the first to be reduced.
According to Legg, the impact will not be uniform. Digital-heavy roles that depend on language, knowledge, mathematics, coding and abstract problem-solving are likely to feel the pressure earlier than others.
AI systems, he noted, already perform strongly in language and general knowledge tasks. Improvements in reasoning, visual understanding and continuous learning are also progressing, bringing machines closer to professional-level performance in many white-collar jobs.
By contrast, work that requires physical interaction with the real world — such as plumbing, electrical work or construction — may remain relatively insulated, at least for now, because automating physical environments is far more complex.
Legg also flagged a larger concern. If machines become consistently better and cheaper at cognitive labour, the traditional economic model — where people exchange mental effort for income — could start to fray.
He warned that dismissing these signals could leave societies unprepared, arguing that early recognition allows time to rethink education, employment structures and income distribution before disruption becomes widespread.
Despite the caution, Legg did not describe the future as bleak. He said AI could unlock a period of higher productivity, scientific advances and economic expansion — what he described as a potential golden age.
The real challenge, he said, will be ensuring that the benefits of this productivity are shared widely, so people retain financial security and a sense of purpose as work itself changes.
Legg stressed that these shifts are unlikely to happen suddenly, but could accelerate quickly once AI systems reach reliable, professional-level capability in knowledge-based tasks.
