12k Workers Survey: Can AI Really Help Companies Reduce Costs and Improve Efficiency?

BCG "AI at Work" report surveyed nearly 12k workers and found that over 40% of non-managerial white-collar employees who regularly use AI tools save an entire day of work each week. But nearly half of the respondents said that the time spent managing and directing AI has exceeded the time spent doing the work itself.
(Background: AI job panic! Microsoft executives warn: Most white-collar workers will be automated out of their jobs "within the next 12-18 months")
(Additional context: Zuckerberg creates "CEO-specific AI agents," with Meta's 78,000 employees starting to let AI agents "socialize" with each other)

Table of Contents

Toggle

• 70% of white-collar workers worldwide are using AI, but results vary
• Managing AI takes more time than doing the work: the "Joy Paradox" emerges
• This is not a technology problem, but a leadership problem

The latest "AI at Work" report from Boston Consulting Group (BCG) shows that over 40% of non-managerial white-collar employees who regularly use AI tools save more than a full workday each week. This number sounds like the dawn of a management revolution, but BCG then states that most companies still don’t know how to turn the time saved into real business value.

70% of white-collar workers worldwide are using AI, but results vary

The survey behind the BCG report is substantial: spanning 14 countries and regions, involving nearly 12k workers, covering four dimensions—AI adoption rate, employee expectations, leadership, and organizational transformation.

The data shows: 74% of non-managerial white-collar workers say they are regular users of AI tools, a significant increase of 23 percentage points from a year ago. This means that in less than 12 months, AI use has shifted from a few early adopters to a daily habit for most white-collar workers.

However, explosive growth in adoption has not simultaneously boosted organizational performance. BCG explicitly states in the report that many companies face a gap between efficiency gains driven by AI and actual business results.

Another survey, involving 3,200 corporate managers by Workday, provides more specific figures: 85% of employees save 1 to 7 hours per week using AI, but nearly 40% of that saved time is immediately eaten up by "rework." Simply put, AI does the work, but you still have to go back and fix things.

Regional differences are also significant. BCG points out that AI usage among non-managerial employees is higher than the global average in India, Brazil, and South Africa, while the US, France, and Italy fall below the average. This distribution contradicts the common intuition about "leading countries in AI application."

Managing AI takes more time than doing the work: the "Joy Paradox" emerges

The most thought-provoking figure in the BCG report is this: nearly half of respondents say they spend more time managing and commanding AI than actually completing their work.

There is a structural contradiction behind this. AI tools indeed make work different—about two-thirds of regular AI users say AI has increased their job satisfaction; but at the same time, 41% say their cognitive load has increased. In simple terms, the mental workload has grown, not decreased.

Sylvain Duranton, co-author of the report and BCG’s expert, calls this phenomenon the "Joy Paradox": AI makes work better and more enjoyable, but also more difficult. He further explains:

"In the early stages of AI use, the novelty and cognitive extension bring joy, but without clear strategy, this 'AI honeymoon' will fade."

This suggests that simply promoting usage rates is not enough. Companies might sustain morale with novelty in the first year of AI adoption, but without systematic process redesign, AI quickly shifts from a solution to another management burden.

Another noteworthy signal is the rise of AI agents capable of autonomously completing complex tasks and connecting multiple tools. BCG reports that 30% of respondents have integrated AI agents into their daily workflows, more than double the percentage from a year ago. Over 60% believe that AI agents will be able to independently complete at least half of their work within three years.

This is not a technology problem, but a leadership problem

Vinciane Beauchene, co-author of the BCG report, states directly:

"Everyone talks about AI replacing jobs, but the real issue is rethinking the added value humans bring to the process. That’s the leader’s responsibility."

This statement hits the core of the current AI application dilemma. Companies invest heavily in deploying AI tools but do not simultaneously redesign workflows or clarify which tasks should be delegated to AI and which require human judgment. Efficiency is unlocked, but no one tells employees where to focus their time.

Multiple surveys show that about 80% of companies see no real profit improvement after AI investments. Economists refer to this phenomenon as the modern "Solow Paradox": technological progress does not seem to translate into productivity gains. The last major wave was during the proliferation of personal computers in the 1980s and 1990s. It took nearly a decade for companies to learn how to truly leverage the efficiency brought by computers.

BCG’s report does not claim that AI’s efficiency gains are false. It states that: efficiency is real, but turning that efficiency into value requires a different set of capabilities—capabilities that most companies currently lack.