Tech leaders say AI means less work – their staff say they work up to 90 hours a week.

For years, a consistent refrain from the titans of the technology industry, those pouring hundreds of billions annually into the development of artificial intelligence, has been that this transformative technology would ultimately liberate humanity from the grind, ushering in an era of reduced working hours and enhanced leisure. This optimistic vision has been painted repeatedly, suggesting a future where AI handles the mundane, leaving humans free for more creative or less strenuous pursuits.

The promises have been specific and bold. As far back as four years ago, an engineering director at Google confidently predicted that AI would pave the way for a widespread four-day work week by the year 2025. This was not a fringe opinion but a sentiment echoing through the industry, fuelled by the rapid advancements in machine learning and automation. The idea was that AI’s ability to automate complex tasks, from data analysis to coding, would drastically cut down the time required for traditional corporate functions, making a shorter work week not just desirable, but entirely feasible.

Earlier this year, and notably just one year shy of Google’s 2025 prediction, OpenAI, a company at the very forefront of AI innovation, took up this mantle in a more formal capacity. They actively urged companies across various sectors to begin trialling a four-day work week, explicitly stating that it should come with no reduction in pay. OpenAI’s rationale was clear: AI’s burgeoning capabilities would soon accelerate human labor to such an extent that the corporate world needed to proactively prepare for a paradigm shift in how work is structured. This advocacy positioned AI not just as a tool for efficiency, but as a catalyst for a more balanced and humane working culture.

However, the reality within these very companies appears to starkly contradict their public pronouncements. A former OpenAI technical employee, who departed the company last year, revealed to the BBC a starkly different internal picture. This individual described a work environment far removed from the promised four-day week, instead characterized by often "gruelling" demands. The company, despite its external advocacy, never actually trialled the shorter work week for its own staff during this employee’s tenure. Instead, the culture was marked by frequent "crisis meetings," an expectation of working weekends, and "super cut-throat" performance reviews that could see colleagues abruptly dismissed. The former employee recounted routinely working on Saturdays or Sundays simply "to catch up or make sure things aren’t broken," illustrating a pervasive sense of urgency and relentless pressure.

Other prominent tech companies have similarly championed the notion of AI dramatically reducing the necessity for human labor. Anthropic, another leading AI firm, proudly showcased its popular coding tool and chatbot, Claude, boasting that it could operate autonomously for seven hours without a break – effectively performing a full corporate workday on its own. Mark Zuckerberg, CEO of Meta, proclaimed that his company was experiencing a pivotal year where "AI starts to dramatically change the way that we work," enabling a significantly smaller workforce to achieve more than ever before. This sentiment, however, coincided with Meta laying off one in ten of its employees, raising questions about whether "doing more with less" truly meant less work overall, or simply less work for fewer people. Anthropic’s chief executive, Dario Amodei, further amplified these concerns by warning that as AI tools inevitably become more productive, the consequence could be even broader job losses across industries.

Despite these claims of AI-driven efficiency and reduced workloads, the very individuals developing and utilizing these advanced AI tools within these tech giants report a vastly different experience. Far from clocking out early, they are consistently working well beyond the traditional five-day, 40-hour work week. While tech workers in places like the US Bay Area are known for being exceptionally well-compensated, and industries like investment banking or law are notorious for long hours, a 14-hour workday was not always the norm in tech. For many years, the sector offered a more typical 9-to-5 office job, attracting talent with promises of innovation and a dynamic, albeit often demanding, environment.

The former OpenAI employee, for instance, reported routinely putting in at least 70 hours a week – a significant increase compared to their previous tech roles. While now working at a new AI-focused startup, they note an improvement in work-life balance, with hours closer to 50-60 a week, "outside of sprints." The term "sprint" is common in technology, referring to intense periods of focused work leading up to a product or feature release. However, within companies like OpenAI and Anthropic, or larger tech firms heavily invested in urgent AI projects, these sprints can be exceptionally extreme, extending for many weeks and pushing work hours past 90 in a single seven-day period, according to tech workers who spoke to the BBC. Neither OpenAI nor Anthropic responded to requests for comment on these intense working conditions.

At Meta, employees described being abruptly "drafted" onto urgent AI projects this year, a term used because they were given no choice in the matter. "They just move you over," a former employee explained, adding, "You can’t say no – or if you do, you have to quit." While Meta has reportedly begun to relax this strict policy, many workers had already been compelled into these demanding AI roles. The hours on such teams at Meta are consistently long, with staff frequently working late into the night and over weekends, often feeling "on call" even during their non-working hours, according to current and former employees.

It’s important to note the legal context: in the US, there are no federal limits on the number of hours a person over the age of 16 can work, creating a permissive environment for such demanding schedules. This contrasts sharply with regulations in the UK and Europe, where laws typically limit the working week to 48 hours, including overtime, aiming to protect worker well-being.

Meta’s current AI projects are ambitious, including the development of an AI tool for software engineering and other tasks, as well as building complex AI model infrastructure to measure the success of AI in replicating human tasks. "You’re literally working in teams of people trying to replicate humans doing jobs," the former Meta employee stated, describing the work as seemingly "endless." A Meta spokesperson declined to comment on these specific claims.

The ripple effect of AI’s ascendance isn’t limited to those directly developing the tools. Even tech workers whose roles are not primarily focused on AI development are experiencing increased workloads. Amin Shali, a former Google employee, left the company in May partly due to the negative impact AI was having on his job. He recounted frequently having to work through the night because internal engineering functions were failing, a problem he attributed to Google diverting crucial resources, such as processing units and memory storage, to its burgeoning AI projects. A Google spokesperson also declined to comment on Shali’s claims. Since leaving Google, Shali reported a significant improvement in his sleep and overall health, concluding that heavy reliance on AI tools in large tech companies "creates a bad culture with excess pressure on engineers."

Emerging academic research further supports the notion that AI tools, rather than reducing burdens, are intensifying them. A study conducted by UC Berkeley, which tracked hundreds of workers at a US tech company over eight months as they integrated AI into their workflows, found a consistent trend: employees "worked at a faster pace, took on a broader scope of tasks, and extended work into more hours of the day." The findings suggest a paradoxical outcome where AI, designed to enhance productivity, effectively expands the parameters of work itself.

Neil Thompson, an innovation scholar at MIT, offers an explanation for this phenomenon. He suggests that even within leading AI development companies, the deployment of these tools doesn’t simply mean workers offload tasks and move on. Instead, any potential time savings are quickly "sucked up by the changes, and implementing them, and making sure they worked." The UC Berkeley research specifically highlighted that the workload of tech employees expanded partly due to the constant necessity of checking and refining the output of AI tools, which are not yet infallible. Furthermore, Thompson points out a human tendency: once processes are smoothed out and time is ostensibly saved by AI, people tend to fill that newly available time with more work. This can be driven by personal ambition, a desire to prove value to employers, or simply the emergence of new, AI-related tasks.

"People assume that 20% less work means four-day weeks," Thompson observed. "But new work emerges." This new work often involves prompt engineering, overseeing AI models, debugging AI-generated code, integrating AI outputs into complex systems, or training AI to perform even more sophisticated tasks. The scope of "work" doesn’t shrink; it evolves and often expands to encompass the management and optimization of the AI itself.

As Amin Shali profoundly noted, "AI is supposed to be doing so much for us now, so many more people should at least have better health and better sleep." Instead, the current reality for many at the forefront of AI development appears to be the opposite: an unrelenting pace, extended hours, and a growing sense of pressure, casting a shadow over the industry’s grand promises of a future liberated by artificial intelligence.