AI Promises Less Work, but Tech Employees Report 70- to 90-Hour Weeks

Key Highlights

• AI and tech workers report working 70 to 90 hours a week.
• Former employees describe weekend work, crisis meetings, and high-pressure performance reviews.
• AI-focused product sprints can continue for weeks.
• Some Meta employees said they were reassigned to urgent AI projects with little choice.
• Research suggests AI can increase work intensity instead of reducing it.
• Employees may use time saved by AI to take on more tasks rather than work fewer hours.

Introduction

AI has been sold as a productivity revolution that could eventually reduce the amount of time people spend working. Some technology leaders have even suggested that AI could make four-day workweeks realistic by automating repetitive tasks and allowing employees to accomplish more in less time.

Inside major AI and technology companies, however, some workers describe a sharply different reality. Employees and former employees have reported 70-hour weeks, extended product sprints, weekend work, late-night assignments, and workloads that can exceed 90 hours during intense development periods.

The contrast raises a fundamental question about workplace automation: does AI actually give people more free time, or does it simply allow companies to demand more output?

AI Companies Have Promised Shorter Workweeks

For years, technology executives have argued that AI could reduce the amount of human labor required for many tasks.

Predictions have included shorter workweeks and significant reductions in routine work. Some companies have even encouraged employers to experiment with four-day schedules as AI tools become more capable.

The logic is straightforward. If AI can write code, analyze documents, summarize information, automate administrative work, and complete other tasks faster than humans alone, employees should theoretically need fewer hours to produce the same amount of work.

In practice, that productivity gain does not always translate into shorter schedules.

Former OpenAI Employee Reports 70-Hour Weeks

One former OpenAI technical employee described a work culture built around frequent crisis meetings, weekend work, and highly competitive performance expectations.

The former employee said they regularly worked at least 70 hours per week, significantly more than in previous technology jobs. After leaving, they moved to another AI startup where they said their typical schedule fell closer to 50 to 60 hours outside intense product-development periods.
The experience illustrates one of the central contradictions of the current AI boom. Companies are building technology designed to automate work while simultaneously asking the people building those systems to work exceptionally long hours.

AI Sprints Can Reach 90 Hours a Week

Product development in technology companies often involves temporary periods known as sprints, when teams work longer hours before a major release.

At AI companies, those sprints can reportedly become much more intense.

Workers described development cycles lasting several weeks in which employees at OpenAI and Anthropic could work more than 90 hours during a seven-day period.

A 90-hour week translates to nearly 13 hours of work every day if spread evenly across seven days.

Even when such schedules are temporary, repeated sprints can create a work culture in which unusually long hours become normal.

Meta Workers Describe Being “Drafted” Into AI Teams

Pressure is not limited to companies focused exclusively on artificial intelligence.

At Meta, current and former employees described being abruptly reassigned to teams working on urgent AI projects. Some employees referred to the process as being “drafted” because they said they had little control over the reassignment.

Workers on those teams reportedly worked late into the night and on weekends while feeling effectively on call even during hours when they were not formally working.

The situation illustrates how the race to develop competitive AI systems can change workplace expectations across entire companies, not just specialized research divisions.

AI Development Has Become an Industry Arms Race

The pressure reflects the enormous competitive stakes surrounding artificial intelligence.

Major technology companies are investing heavily in AI infrastructure, computing capacity, research, and product development. Each company wants to release more capable models, improve software tools, and establish a strong position before competitors gain an advantage.

That urgency can translate directly into employee workloads.

When companies view AI as a once-in-a-generation technological shift, managers may treat projects as permanently urgent. Temporary product sprints can then become a recurring operating model.

Even Non-AI Workers Can Feel the Pressure

Employees do not necessarily need to work directly on AI systems to experience the consequences.

A former Google employee said internal engineering functions became less reliable as critical computing resources such as processing capacity and memory were redirected toward AI projects.

He described frequently working late into the night to address technical failures and said his sleep and overall health improved after leaving the company.

This suggests AI investment can affect workloads throughout an organization by changing how infrastructure, budgets, and engineering resources are allocated.

Research Suggests AI Can Intensify Work

The experience described by workers is consistent with emerging research on workplace AI.

A UC Berkeley study that followed hundreds of employees at a U.S. technology company for eight months found that workers using AI moved faster, handled a wider range of responsibilities, and extended their work across more hours of the day.

Instead of shrinking workloads, AI may therefore increase the amount of work employees can realistically be expected to complete.

Greater productivity can become a new baseline rather than a source of additional free time.

AI Output Still Requires Human Supervision

Another reason AI may not reduce working hours is that automated systems still require significant human oversight.

Employees must check AI-generated output, identify errors, correct problems, update workflows, and make sure automated systems perform reliably.

The UC Berkeley research found that constantly verifying AI-generated work contributed to expanding employee workloads.

That means automation can create new categories of work even as it eliminates others.

An employee may spend less time producing a first draft or writing initial code, but more time reviewing, testing, integrating, and monitoring the output.

Productivity Gains Can Simply Create More Work

One of the most important dynamics is what happens after AI actually saves time.

Organizations rarely leave that newly available time unused. Employees may receive more assignments, broader responsibilities, or higher performance expectations.

MIT innovation scholar Neil Thompson described how time savings can be absorbed by new tasks, system changes, and the work required to ensure AI tools function correctly.

This creates a productivity paradox.

If AI makes an employee 20% faster, the company may not reduce that employee’s schedule by 20%. It may instead expect 20% more output.

The Four-Day Workweek May Not Arrive Automatically

The idea that productivity gains naturally produce shorter workweeks has historical appeal, but workplace incentives often operate differently.

Companies generally want to maximize productivity, growth, and competitive advantage. Employees may also feel pressure to demonstrate their value, especially when automation raises questions about which jobs remain necessary.

As a result, workers may voluntarily fill AI-created time savings with additional work because they want to remain competitive or fear appearing less productive.

The four-day workweek therefore requires deliberate organizational decisions. AI alone does not guarantee it.

Job Security Adds Another Layer of Pressure

AI-driven productivity also creates uncertainty around employment.

Technology executives have warned that increasingly capable AI systems could allow smaller teams to produce the same amount of work, potentially reducing the number of employees companies need.

That can create powerful incentives for remaining employees to work harder.

If workers believe AI may eventually replace part of their role, they may feel pressure to demonstrate greater output, learn new tools faster, and take on more responsibilities.

In this environment, AI can simultaneously promise greater efficiency and increase anxiety about job security.

Long Hours Could Create a Sustainability Problem

The current pace of AI development may be difficult to maintain indefinitely.

Repeated 70- to 90-hour weeks can increase burnout, employee turnover, mistakes, and health problems. Companies competing for highly specialized engineers may ultimately discover that extreme workloads make retaining talent more difficult.

This creates another paradox for the AI industry.

The companies developing tools designed to increase human productivity may need to rethink how they measure productivity inside their own organizations.

What Companies Need to Decide

The central issue is not whether AI saves time. In many tasks, it clearly can.

The more important question is what organizations choose to do with those savings.

Companies could use AI productivity gains to reduce hours, improve work-life balance, and maintain the same output with less human effort.

Or they could use the technology to increase targets, accelerate release schedules, reduce staffing, and demand more output from every employee.

The technology itself does not determine which outcome wins.

Conclusion

Artificial intelligence may eventually transform the traditional workweek, but the early experience inside some of the companies leading the AI revolution suggests that greater productivity does not automatically mean less work.

Workers have described schedules reaching 70 to 90 hours a week, extended product sprints, weekend assignments, and relentless pressure to build and deploy AI systems faster. Research also suggests that AI can expand workloads by increasing expectations and creating new responsibilities around verification and implementation.

The AI productivity revolution may therefore create a surprising workplace challenge: technology can give employees more time, but employers still decide whether workers get to keep it.