What is the story about?
As companies encourage employees to use artificial intelligence tools to ease routine tasks and boost productivity, new research suggests the effects may be more complex than anticipated, according to the
Harvard Business Review.
An in-progress eight-month study at a US-based technology company with about 200 employees found that generative AI tools did not reduce workloads. Instead, they accelerated the pace of work, broadened job responsibilities and extended working hours, often without any formal requirement from management. The company offered enterprise subscriptions to commercially available AI tools but did not mandate their use.
Researchers observed work in person two days a week, monitored internal communication channels and conducted more than 40 in-depth interviews across engineering, product, design, research and operations between April and December last year.
One key pattern identified was task expansion. Because generative AI tools could bridge gaps in knowledge, employees increasingly took on responsibilities that previously belonged to others.
Product managers and designers began writing code, researchers undertook engineering tasks, and staff across departments attempted work they might earlier have outsourced or avoided. AI tools provided immediate feedback and correction, which many workers described as empowering and confidence-building. Over time, these small experiments expanded job scopes and absorbed work that might otherwise have required additional hiring.
This shift also created additional demands. Engineers spent more time reviewing and correcting AI-assisted work and advising colleagues who were completing partially finished pull requests. Much of this oversight took place informally through Slack messages or brief consultations, adding to workloads.
The study also found that AI reduced the friction of starting tasks, making it easier for employees to work during moments previously reserved for breaks. Some prompted AI tools during lunch, in meetings or before leaving their desks so that tasks could progress while they were away.
Although these actions felt minor, over time they reduced natural pauses in the working day. Because prompting AI systems often resembled informal conversation, work gradually extended into evenings and early mornings without deliberate planning.
Several employees reported that downtime no longer provided the same sense of recovery. Work became more continuous, with boundaries between professional and personal time easier to cross.
Generative AI also altered the rhythm of work. Employees frequently managed multiple threads simultaneously, writing code while AI generated alternatives, running parallel agents or reviving deferred tasks in the background.
While this created a sense of momentum, it also required frequent shifts of attention and monitoring of outputs. The result was a growing number of open tasks and increased cognitive load.
Over time, faster workflows led to higher expectations for speed. Although automation was intended to save time, many workers reported feeling busier and under greater pressure than before.
As one engineer said, “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.”
Researchers described a self-reinforcing cycle in which AI accelerated tasks, raised expectations and increased reliance on the technology. As employees attempted a wider range of work, the overall density of tasks grew.
Although the expansion of work was voluntary and often experienced as enjoyable experimentation, the study warned that it could conceal rising cognitive strain. Short-term productivity gains may mask fatigue, burnout and weakened judgement over time.
Employees reported feeling productive but not less busy, and in some cases busier than before.
The findings suggest organisations should develop structured norms around AI use, described as an “AI practice”. This would define how and when AI should be used, establish boundaries for workload expansion and create deliberate stopping points.
One proposed measure is the introduction of intentional pauses. Brief, structured intervals could allow workers to reassess decisions, question assumptions or reconnect tasks to organisational goals before proceeding.
Such pauses are not intended to slow progress but to prevent unchecked acceleration and help sustain healthier, more durable productivity in AI-supported environments.
An in-progress eight-month study at a US-based technology company with about 200 employees found that generative AI tools did not reduce workloads. Instead, they accelerated the pace of work, broadened job responsibilities and extended working hours, often without any formal requirement from management. The company offered enterprise subscriptions to commercially available AI tools but did not mandate their use.
Researchers observed work in person two days a week, monitored internal communication channels and conducted more than 40 in-depth interviews across engineering, product, design, research and operations between April and December last year.
Task expansion across roles
One key pattern identified was task expansion. Because generative AI tools could bridge gaps in knowledge, employees increasingly took on responsibilities that previously belonged to others.
Product managers and designers began writing code, researchers undertook engineering tasks, and staff across departments attempted work they might earlier have outsourced or avoided. AI tools provided immediate feedback and correction, which many workers described as empowering and confidence-building. Over time, these small experiments expanded job scopes and absorbed work that might otherwise have required additional hiring.
This shift also created additional demands. Engineers spent more time reviewing and correcting AI-assisted work and advising colleagues who were completing partially finished pull requests. Much of this oversight took place informally through Slack messages or brief consultations, adding to workloads.
Boundaries between work and rest blurred
The study also found that AI reduced the friction of starting tasks, making it easier for employees to work during moments previously reserved for breaks. Some prompted AI tools during lunch, in meetings or before leaving their desks so that tasks could progress while they were away.
Although these actions felt minor, over time they reduced natural pauses in the working day. Because prompting AI systems often resembled informal conversation, work gradually extended into evenings and early mornings without deliberate planning.
Several employees reported that downtime no longer provided the same sense of recovery. Work became more continuous, with boundaries between professional and personal time easier to cross.
Rise in multitasking and cognitive load
Generative AI also altered the rhythm of work. Employees frequently managed multiple threads simultaneously, writing code while AI generated alternatives, running parallel agents or reviving deferred tasks in the background.
While this created a sense of momentum, it also required frequent shifts of attention and monitoring of outputs. The result was a growing number of open tasks and increased cognitive load.
Over time, faster workflows led to higher expectations for speed. Although automation was intended to save time, many workers reported feeling busier and under greater pressure than before.
As one engineer said, “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.”
Risks of silent workload creep
Researchers described a self-reinforcing cycle in which AI accelerated tasks, raised expectations and increased reliance on the technology. As employees attempted a wider range of work, the overall density of tasks grew.
Although the expansion of work was voluntary and often experienced as enjoyable experimentation, the study warned that it could conceal rising cognitive strain. Short-term productivity gains may mask fatigue, burnout and weakened judgement over time.
Employees reported feeling productive but not less busy, and in some cases busier than before.
Call for structured ‘AI practice’
The findings suggest organisations should develop structured norms around AI use, described as an “AI practice”. This would define how and when AI should be used, establish boundaries for workload expansion and create deliberate stopping points.
One proposed measure is the introduction of intentional pauses. Brief, structured intervals could allow workers to reassess decisions, question assumptions or reconnect tasks to organisational goals before proceeding.
Such pauses are not intended to slow progress but to prevent unchecked acceleration and help sustain healthier, more durable productivity in AI-supported environments.
