Will AI Bring a 3-Day Workweek? Tech Employees Reveal Weekly Hours Exceeding 90 Hours

(CNA, San Francisco, 10th - international news compilation) In recent years, tech giants have invested hundreds of billions of dollars annually in developing various artificial intelligence (AI) tools. Company executives insist that AI will ultimately reduce human working hours. However, tech industry employees report that their workloads have instead increased, with some working over 90 hours per week during peak periods.

According to a BBC report, a Google engineering director predicted four years ago that AI would enable a 'three-day weekend' by the end of 2025.

Earlier this year, OpenAI partially embraced this challenge. The company formally urged businesses to begin trialing a four-day workweek with no pay cut, claiming that AI would soon dramatically boost human productivity and that companies should prepare accordingly.

However, a former technical employee at OpenAI, who left last year, told BBC that during their tenure, the company never actually implemented the very four-day workweek policy it advocated.

The former employee described OpenAI's work culture as extremely harsh, with frequent 'crisis meetings,' weekend overtime, and 'extremely brutal' performance reviews being commonplace—employees could be suddenly fired as a result.

Anthropic also emphasized that its popular programming tool and chatbot, Claude, can operate autonomously for seven consecutive hours without rest—equivalent to a full corporate workday.

Meta CEO Mark Zuckerberg stated that the company is in a period where 'AI is dramatically changing how we work,' and that these tools allow fewer employees to accomplish more than ever before.

Meta has already cut about one-tenth of its workforce, and Anthropic CEO Dario Amodei warned that as AI tools become more effective, job losses could intensify.

Despite this, employees at tech companies—both developing and using AI tools that are said to replace at least some human tasks—report that their working hours far exceed the traditional five-day, eight-hour workweek.

● Workloads Exceeding 90 Hours Per Week During 'Sprints'

For years, the U.S. tech industry leaned toward a 9-to-5 office culture, with 14-hour workdays not being the norm. But that has changed.

A former OpenAI employee said they worked at least 70 hours per week—far exceeding their previous roles in other tech jobs. This individual now works at another AI-focused startup, where they work 50 to 60 hours per week outside of 'sprint' periods.

'Sprints' are a common term in the tech industry, referring to the weeks leading up to the launch of a new feature or product, during which employees must work long hours under intense pressure.

Tech employees interviewed by BBC revealed that sprints at OpenAI and Anthropic can last for several weeks, with some weeks exceeding 90 hours of work in seven days.

● Employees 'Conscripted' into AI Teams

Former and current Meta employees reported being abruptly reassigned this year to other teams to work on AI-related tasks deemed urgent. They referred to these arrangements as 'conscription' because employees had no choice.

The former employee said: 'You can't say no, or you have to resign.'

They added that although Meta has recently begun to soften its 'take it or leave it' approach, many employees have still been forced to change roles.

According to current and former employees, work hours in these Meta teams are often long, with staff frequently working late into the night or on weekends. Even outside working hours, employees feel they must remain 'on call.'

Even tech employees not directly involved in AI tool development are experiencing increased workloads due to AI.

Amin Shali, a former Google employee, said he left in May because he believed AI had negatively impacted his work—for example, when internal engineering functions failed, he often had to work through the night.

As Google shifts critical resources like processors and memory to AI projects, such incidents are becoming more frequent.

Shali told BBC that after leaving, his sleep and overall health improved significantly. He realized that large tech companies' heavy reliance on AI tools 'created a culture of excessive exploitation of engineers.'

● Time Saved by AI Is Consumed Elsewhere

New research suggests that AI tools may actually increase employee workloads.

A University of California, Berkeley study tracked hundreds of employees at a U.S. tech company over eight months, observing their use of AI. The study found that these employees 'worked faster, took on broader tasks, and worked longer daily hours.'

Neil Thompson, an innovation scholar at MIT, said that even tech companies leading AI development are unlikely to tell employees they can go home after using AI to complete their tasks.

Thompson said: 'Even if time is genuinely saved, that time is consumed by modifying systems, implementing changes, and ensuring they function properly.'

The UC Berkeley study also found that tech employees' workloads increased partly because they must constantly verify AI tool outputs.

Thompson said: 'People think reducing 20% of the workload means a three-day weekend, but new tasks keep emerging.' (Translation: Hung Pei-ying) 1150810