In many workplaces around the world, a strange shadow looms behind the glow of progress. Artificial intelligence, once heralded as the tool that would free workers from tedious tasks, is increasingly becoming a source of stress, anxiety, and burnout. The disconnect between what employers expect of AI and how workers experience it is threatening to reshape labor markets in ways that go beyond job automation.
The Tension Between Promise and Practice
AI is being sold as a productivity booster, a tool to reduce routine burdens, to accelerate decision making, to allow workers to focus on more creative or strategic tasks. Some of this is happening. Companies are investing heavily; reports show nearly all companies plan to increase AI investments over the next few years.
But many workers say the promise hasn’t materialized, at least not the way it was pitched. In surveys:
Upwork found that 77% of full-time employees say AI tools increased their workload or decreased their productivity.
Nearly half of workers using AI say they don’t even know how to meet their employer’s expectations about productivity gains through AI.
In another survey, 71% report being burned out; 65% say they’re struggling with the heightened productivity demands.
So the tension is twofold: employers often assume AI will deliver efficiency gains automatically and immediately; employees, often lacking training, clarity, or support, find themselves juggling new responsibilities, new tools, and often more scrutiny over output especially when errors or “hallucinations” in AI outputs require human oversight.
Employee Readiness and the Learning Gap
An important contributor to the friction is the gap in readiness. Readiness includes not just access to tools, but:
Skill levels: How many are confident using AI? Studies show a wide gap between what leadership believes and what employees feel. In McKinsey’s “Superagency” report, for example, only 1 percent of companies believe they are mature in their deployment of AI. And employees are already using AI more regularly than many leaders think.
Clarity of expectations: Many employers expect certain productivity gains but are not specifying what “good use” of AI looks like. This leaves employees in the dark, sometimes literally unclear how to meet targets or even how to use the tools well.
Support structures: Training programs, change management, mental health support, and transparent policies are often under-developed. In many cases, systems are being retrofitted to include AI rather than designed around human needs.
Because of these gaps, many workers report feeling overwhelmed: learning to use new tools, checking work generated by AI for errors, keeping up with rapid updates, and having less margin for mistakes. All of it contributes to what some commentators are calling “AI fatigue” or “AI exhaustion.”
The Burnout Risks and Reality
Burnout isn’t a distant risk; it’s happening now. The evidence:
Job burnout rates are rising: in 2025, one study showed 66% of workers reported burnout.
A substantial number of employees indicate they may leave their jobs soon because of stress related to being overworked or overwhelmed.
Mental health impacts are real: anxiety over job loss, constant pressure to upskill, fear of falling behind, or being replaced, all of which can exacerbate stress.
In some sectors, the “always-on” expectation is also growing: because AI can run 24/7, because collaboration tools blur boundaries, workers feel less able to disconnect.
What This Means for the Labor Market
These dynamics will likely ripple through labor markets in several ways:
Polarization of skill levels and wage gaps:
Workers who can adapt, who have strong technical and adaptive skills, may benefit substantially. Those who can’t, or who lack access to training, may fall further behind, possibly relegated to less secure or lower-paid roles. The divide between “augmented” and “displaced” tasks will widen.Changing role structures and job design:
Entry-level jobs may change fundamentally. Tasks once seen as routine may vanish; expectations for multi-tasking, tool use, and rapid learning become built into many roles. Job descriptions will shift. In some cases, AI may allow reduced hours or compressed workweeks, if managed well.Employee retention and turnover challenges:
If workers feel overwhelmed, or that expectations are unfair or unreasonably high, companies may see more turnover. That has costs; recruiting, training, loss of institutional knowledge, morale damage. Organizations that do not manage the human side of AI adoption risk losing more than they gain by adding tools.Pressure on education, training, and policy:
Governments, education institutions, and firms will be under increasing pressure to supply upskilling, continuous education, and flexible training models. Lifelong learning may move from “nice-to-have” to essential. Policy may need to catch up, regulating transparency, protecting workers, perhaps even rethinking labor standards around workloads.Potential adjustments in employer expectations and practices:
To sustain AI adoption without burning out the workforce, organizations may need to rethink how they measure productivity (not just quantity but quality, balance, human oversight cost), how they roll out tools (incrementally, with feedback), how they support workers (training, mental health), and possibly how they redistribute gains (do fewer people working more, or many people sharing work more efficiently). Some suggest the idea of designating tasks less suitable for automation (creative, interpersonal, or emotionally taxing work) to ensure human engagement, reducing the risk of isolation.
Where We Go From Here
There is no universal path forward, but a few guiding principles seem clear:
Transparency and involvement: Companies should be open about what AI tools will do, what expectations will change, and involve workers in that design process.
Training plus patience: Investment in training isn’t enough if it’s rushed or superficial. Employees need time to master tools, adjust workflows, and fail safely.
Redefining “productivity gains”: AI productivity gains should account for the hidden costs; oversight, error correction, learning curves, mental health. Unrealistic targets based only on potential gains invite burnout.
Work design that protects human elements: Even as AI takes over routine tasks, corporate culture, collaborative work, mentoring, creativity, judgment are areas where human work remains essential, and where support is needed to prevent alienation.
Policy support: Labor law, employment standards, safety net programs, and social dialogue (unions, worker representation) will need to address the AI era’s particular challenges.
Conclusion
AI is already reshaping work. What remains to be seen is whether this shift will lead to a more empowered, efficient workforce, or a fatigued, overburdened one. In theory, AI could free workers for more meaningful tasks; in practice, many feel they are doing more than before, not less. The outcome may depend less on the prowess of the algorithms than on the quality of the human systems that surround them: how employers set expectations, how workers are trained and supported, and how we define success in an age of swift technological change.