AI Doesn't Reduce Work. It Intensifies It.

The productivity paradox hiding inside your AI rollout

There is a widely held assumption driving enterprise AI adoption right now: that AI tools will reduce the burden of routine work, free up cognitive bandwidth, and allow employees to focus on higher-value tasks. Leaders are banking on this premise. Many of them may be in for a rude awakening.

New research from an eight-month study of a US-based technology company reveals a more complicated and frankly more alarming picture. AI didn't reduce work. It intensified it. And the organisations scrambling to get more employees using AI may not see the consequences until the damage is already done.

The study, conducted from April to December, involved in-person observation two days per week, monitoring of internal communication channels, and deep-dive interviews across multiple disciplines. The company offered enterprise subscriptions to commercially available AI tools but imposed no requirement to use them. Workers adopted AI on their own initiative. And that is precisely what makes the findings so significant.

Three ways AI intensifies work

The researchers identified three primary mechanisms through which AI use expanded, rather than contracted, the volume and density of work. Understanding each of them is essential for any organisation currently building out its AI strategy.

Task expansion: doing more because you can

When AI fills knowledge gaps, workers stop deferring tasks they previously would have outsourced or avoided entirely. Product managers and designers began writing code. Researchers took on engineering responsibilities. Individuals across the company attempted work that would previously have justified additional headcount or specialist input.

This felt empowering in the moment. Workers described it as "just trying things" with AI assistance. But those experiments accumulated into a meaningful widening of job scope. And crucially, the extra load rarely came with extra recognition or resource.

"You had thought that maybe, oh, because you could be more productive with AI, then you save some time, you can work less. But then really, you don't work less. You just work the same amount or even more." , Engineer, study participant

There were knock-on effects too. Engineers found themselves spending more time reviewing, correcting, and coaching colleagues who were vibe-coding their way through pull requests using AI tools. This informal oversight surfaced in Slack threads and side conversations, adding to already-stretched workloads without ever appearing in any productivity metric.

Boundary erosion: work that never quite stops

Because AI dramatically reduces the friction of starting a task, workers began filling previously idle moments with small prompts. A quick query during lunch. A follow-up before leaving the desk. An early-morning check of overnight AI outputs. None of these felt like "doing work" in the traditional sense, but collectively they eroded the natural pauses that allow cognitive recovery.

The conversational style of modern AI interfaces made this especially insidious. Typing a prompt to an AI assistant feels nothing like drafting a formal document. It feels like chatting. That perceptual softening made it easy for work to spill into evenings and weekends without any deliberate intention to do so.

Multitasking: the illusion of partnership

AI introduced a new operational rhythm in which workers managed multiple simultaneous threads: writing code manually while an AI generated an alternative version, running several agents in parallel, or reviving long-deferred tasks because AI could "handle them" in the background. Workers described feeling as though they had a capable partner helping them move through their backlog.

In reality, this created a state of near-constant context-switching, frequent checking of AI outputs, and a growing pile of open tasks. Cognitive load increased even as the experience felt productive. Expectations for speed rose, and those expectations became normalised across teams, creating pressure that was structural rather than individual.

<

>

The self-reinforcing cycle no one is talking about

The three mechanisms above do not operate in isolation. They compound each other. AI accelerates certain tasks, which raises expectations for pace. Higher expected pace increases reliance on AI. Greater reliance widens the scope of what workers attempt. A broader scope expands the quantity and density of work. And around it goes.

Several participants in the study explicitly noted that although they felt more productive, they did not feel less busy. Many felt busier than before AI adoption. The promise of reduced burden remained perpetually just out of reach.

Workers consistently expanded their workload voluntarily, not under instruction, because AI made "doing more" feel possible, accessible, and in many cases intrinsically rewarding , study findings, in-progress research

This is where the risks become genuinely serious. What looks like a productivity explosion in the short term can mask silent workload creep and growing cognitive strain. Because the extra effort is framed as enjoyable experimentation, it is easy for leaders to miss how much additional load workers are actually carrying. Over time, the consequences are predictable: fatigue, degraded decision-making, higher error rates, and eventually burnout and turnover.

For more on how AI is reshaping organisational workflows and consumer behaviour across the region, see our coverage of how AI already changed how Asia shops.

The Asia-Pacific Picture

The dynamics uncovered in this research are not limited to US technology firms. Across Asia-Pacific, the same forces are playing out at scale, often in markets where expectations of employee availability and pace are already high.

In Japan, where chronic overwork has been a documented public health concern for decades, the arrival of AI tools that make it easier to work longer hours without obvious fatigue signals raises specific alarm. Japan's government has been pushing for work-style reform legislation since 2018, but AI adoption that quietly erodes boundaries could undermine those gains.

In South Korea, technology sector workers already operate in some of the longest working-hour environments in the OECD. The introduction of AI tools into firms like Samsung, Kakao, and Naver, all of which have accelerated internal AI development and deployment, adds a new layer of complexity. When AI makes expansion feel natural and voluntary, it becomes harder for labour protections to keep pace.

India's IT services sector, which employs millions and serves clients globally, is another critical context. Firms including Infosys, Wipro, and TCS have all announced significant AI integration programmes. If AI tools cause individual contributors to absorb tasks that previously required additional headcount, the implications for hiring, team structure, and employee welfare are substantial.

Meanwhile, Vietnam is taking a notably proactive policy stance on AI. Having recently enforced Southeast Asia's first AI law, Vietnamese regulators are beginning to grapple with the broader socioeconomic implications of AI-driven workplace change. Whether that legislation will address AI-induced workload intensification remains to be seen, but it marks a meaningful start.

China, which has placed AI at the centre of its industrial planning, is also a critical reference point. China's five-year plan places AI at the heart of its next economic phase, prioritising output and capability gains. Whether worker sustainability features in that calculus is an open question, but one that domestic and international observers should be pressing.

Across the region, the pattern is consistent: AI adoption is being driven by productivity ambition, not by any systematic thinking about how work rhythms, cognitive load, and human recovery time will be managed in an AI-augmented environment. That gap is where the real risk lives.

What an "AI practice" actually looks like

The researchers argue that asking employees to self-regulate is not a workable solution. Instead, organisations need to develop deliberate norms around AI use. They call this an "AI practice": a structured set of routines that govern how AI is used, when it is appropriate to pause, and how work should and should not expand in response to new capability.

Three specific interventions are recommended:

1. Intentional pauses. Structured moments that regulate tempo. Before a major decision is finalised, require one explicit counterargument and one clear link to organisational goals. These brief checkpoints do not slow total output. They prevent the quiet accumulation of overload and protect decision quality.
2. Sequencing. Norms that shape when work moves forward, not just how fast. This means batching non-urgent notifications, holding updates until natural breakpoints, and protecting focus windows that shield workers from interruption. When work advances in coherent phases rather than as a continuous stream, fragmentation decreases and context-switching costs fall.
3. Human grounding. As AI enables more solo, self-contained work, organisations need to deliberately protect time for human connection. Brief check-ins, shared reflection moments, and structured dialogue interrupt the individualising effects of continuous AI engagement. They also support creativity by exposing workers to multiple human perspectives that a single AI synthesis cannot replicate.

These are not soft, feel-good additions to a productivity programme. They are structural interventions designed to preserve the conditions under which genuine, sustainable productivity is actually possible.

AI practice vs. passive AI adoption: a comparison

The broader question of how AI is reshaping the labour market connects to education as well. Vietnam's commitment to teaching AI from primary school level reflects an understanding that the next generation of workers will need to engage with these tools differently from the outset. Building healthy AI work habits early may be the most effective long-term intervention available.

For organisations currently navigating AI integration, the findings from this research should prompt a straightforward internal audit: are your productivity metrics capturing the full cost of output gains? Are you measuring hours worked, cognitive strain, and error rates alongside task completion velocity? If not, you may be celebrating a short-term surge that is quietly building a long-term problem.

The broader discourse around AI and the future of work has tended to focus on job displacement. This research suggests a different and perhaps more immediate risk: not that AI will take jobs, but that it will cause the humans who remain to do far more than is sustainable, voluntarily, without ever being asked.

Frequently Asked Questions

Does AI actually increase productivity or just increase workload?

Based on this research, AI increases the volume and pace of work rather than reducing it. Workers completed more tasks and worked longer hours after AI adoption, but reported feeling busier, not less burdened. Genuine productivity gains are possible, but only when organisations actively manage scope, boundaries, and cognitive load through deliberate AI practice norms.

What is AI-induced workload creep and why does it matter?

AI workload creep is the gradual, often invisible expansion of individual job scope and working hours driven by AI tools that make additional tasks feel accessible and low-effort. It matters because it accumulates silently, is framed as voluntary, and is therefore easy for managers to miss until it results in burnout, degraded decision-making, and employee turnover.

What can organisations do to prevent AI burnout among employees?

Researchers recommend building an "AI practice": a structured set of norms governing when and how AI is used. Specific interventions include intentional decision pauses, sequencing protocols that batch notifications and protect focus time, and regular human connection moments that interrupt the individualising effects of continuous AI engagement.

YOUR TAKE

We cover the story. You tell us what it means on the ground.

Related Articles