It would be reasonable—almost intuitive—to assume that the spread of automation and intelligent software would eliminate the most tedious parts of modern labor, freeing workers from drudgery and ushering in a calmer, more humane rhythm to the workday. This assumption, however, quietly ignores one of the most persistent forces in contemporary capitalism: productivity gains are never allowed to sit idle. The lever only moves forward. Any efficiency unlocked by technology is swiftly repurposed, not to grant rest or breathing room, but to justify the introduction of additional tasks, expanded responsibilities, and higher expectations. Relief, when it appears at all, is fleeting.
This dynamic is vividly illustrated in a recent case study drawn from what researchers describe as “in-progress” work by Aruna Ranganathan, a management professor at UC Berkeley, and Xingqi Maggie Ye, a doctoral candidate in Ranganathan’s Berkeley-based program. Their findings suggest that artificial intelligence does not meaningfully simplify the working day. On the contrary, it appears to amplify it. Rather than smoothing workflows or easing cognitive strain, AI tools tend to compress time, intensify output, and accelerate the pace at which workers are expected to perform.
Put more bluntly, the world they describe sounds deeply unpleasant—less a technological utopia than a white-collar pressure cooker.
And yet, for certain corners of the professional world, this is not a nightmare but a feature. If this kind of hyper-accelerated labor environment appeals to you, there’s a decent chance you already work somewhere like Silicon Valley, or at a company steeped in the ideology of relentless optimization. OpenAI itself offers a revealing example. CEO Sam Altman has publicly described how AI has heightened the intensity of his own work, speaking with a mix of wonder and reverence about its effects—even as he shows little visible hesitation about the prospect of large-scale job displacement among knowledge workers. In an interview last October, Altman remarked that he struggles to generate ideas quickly enough to keep pace with what AI now makes possible. The implication, he suggested, is that everything simply moves faster: experimentation, iteration, and decision-making all occur at a breakneck speed.
This sense of being swept along by an accelerating current mirrors the experiences described by employees featured in Ranganathan and Ye’s research, which was summarized in a Harvard Business Review article. The study followed workers at a roughly 200-person company over eight months, examining how generative AI reshaped daily work habits. The authors observed that employees operated at a faster tempo, juggled a wider array of responsibilities, and extended their working hours—often voluntarily, and frequently without explicit managerial direction.
Importantly, this was not an organization that forced AI adoption through top-down mandates. Instead, leadership simply made enterprise-grade AI tools readily available and allowed employees to integrate them organically into their workflows. This was no assembly line operation. The roles in question involved software development, technical problem-solving, and constant digital communication—classic knowledge work. Many of these employees were engineers and developers, quite possibly relying on advanced coding assistants or tools similar to Claude Code to meet growing demands.
As AI became embedded in daily routines, the boundaries between roles began to blur. Workers expanded the scope of their own jobs, stepped into responsibilities once handled by others, and informally assumed positions as mentors, troubleshooters, and quality controllers—coaching colleagues on code, correcting machine-generated output, or refining work produced through so-called “vibe coding.” In many cases, hiring additional staff was delayed or avoided entirely. Employees effectively absorbed workloads that might previously have justified new headcount.
The creep didn’t stop there. Workers reportedly fed prompts into AI systems while sitting in meetings, multitasked with automated tools during brief pauses in their day, and continued submitting requests while waiting for software to load—or even while on lunch breaks. Time that once might have served as a mental reset became another opportunity for output.
How one reacts to this scenario depends heavily on context. In a startup operating in perpetual “founder mode,” where punishing hours are justified by the promise of future equity and the dream of unicorn-level success, this arrangement may sound exhilarating. For founders and executives with billionaire ambitions, an AI-enabled workforce that self-expands its capacity is close to ideal.
But that worldview is hardly universal.
A 2024 Pew Research Center survey paints a more ambivalent picture of worker sentiment in the United States. Roughly half of respondents described themselves as only somewhat satisfied—or outright dissatisfied—with their jobs, while the remaining half reported being very or extremely satisfied. Among lower-income workers, the share who described themselves as highly satisfied dropped from 50 percent to 42 percent.
Notably, the survey found that the most fulfilling aspect of work for most people wasn’t skill accumulation or productivity gains, but human connection. Sixty-four percent of respondents said they were very or extremely satisfied with their relationships with coworkers. By contrast, only 37 percent expressed high satisfaction with opportunities for skill development.
Against that backdrop, it’s difficult to believe that forcing fewer people to do more, pushing work into breaks, and continually expanding job scope will meaningfully improve morale or well-being for the majority of workers—though perhaps this skepticism betrays a lack of the particular “vision” prized in tech leadership circles.
Consider how this scenario looks outside the bubble of software engineering. If you’re a hospital receptionist, a school administrator, or a municipal employee, the promise of AI likely doesn’t translate into thrilling productivity gains. Instead, it may mean delayed hiring, informal role creep, unpaid labor during downtime, and the expectation that you’ll now use complex enterprise AI tools not just to do your job, but to effectively build or maintain your own software solutions.
Even within the tech sector, enthusiasm is far from universal. There’s reason to question whether the productivity gains described in Ranganathan and Ye’s study represent genuine efficiency or simply a well-disguised illusion. An anonymous employee at the cybersecurity firm CrowdStrike wrote to the labor-focused newsletter Blood in the Machine last year, describing a workplace where staff were encouraged to absorb increased per-capita workloads by “working harder and sometimes working longer,” without any corresponding increase in pay. While the company’s machine learning systems were said to perform admirably, the worker expressed deep skepticism that generative AI had actually improved productivity, citing the constant need for proofreading, debugging, and what they called “general babysitting” of AI outputs.
The promised easing of the burden, this employee concluded, never materialized. Instead, workloads grew heavier, expectations rose, and morale sank to historic lows.