AI productivity gains offset by rework costs, study finds

A new global study suggests that while artificial intelligence is boosting employee productivity, organisations are losing a significant portion of those gains to the hidden cost of fixing AI-generated work.

Research released by Workday reveals that nearly 40% of expected productivity improvements from AI are being eroded by time spent correcting low-quality outputs, highlighting a widening gap between AI’s promise and its real-world workplace impact.

The study, conducted in November 2025 by Hanover Research, surveyed 3,200 full-time employees and leaders across North America, Europe, the Middle East and Africa (EMEA), and Asia-Pacific.

While 77% of employees reported increased productivity from AI over the past year, Workday found that approximately 37% of the time saved is being consumed by rework. In practical terms, for every 10 hours of efficiency gained through AI, nearly four hours are lost to correcting, clarifying, or rewriting AI-generated content.

The report described this phenomenon as an emerging drag on AI-driven efficiency.

Heavy users face an “AI tax”

The burden is most visible among frequent AI users. Highly engaged employees spend roughly 1.5 weeks per year fixing AI outputs, a dynamic researchers labelled an “AI tax on productivity.” Notably, only 14% of employees consistently achieve net-positive outcomes from AI use.

Younger workers appear to be disproportionately affected. Employees aged 25 to 34 account for nearly half of those experiencing the highest levels of verification and correction work.

“These employees tend to use AI frequently and with confidence, but they also report spending significantly more time auditing results,” the report noted.

Training gap widens disconnect

The findings also expose a notable misalignment between leadership priorities and employee experience. Although 66% of leaders identified skills training as a top investment priority, only 37% of employees most exposed to AI rework said they had increased access to training. The report further found that fewer than half of organisational roles have been updated to include AI-related skills.

In organisations struggling to realise net productivity gains, less than 25% of roles are considered AI-ready — suggesting many companies have layered AI onto legacy job designs without redesigning workflows.

“AI has been layered onto roles that were never updated to accommodate it,” the study stated.

A model for success emerges

Despite the challenges, the research identified a cohort of employees, dubbed “Augmented Strategists”, who consistently deliver net productivity gains.

Among this group, 93% use AI primarily to identify patterns rather than fully perform tasks, and 79% report receiving increased skills training. Nearly all say they would recommend their organisation as a place to work, pointing to a strong link between AI enablement and employee experience.

Reinvestment priorities under scrutiny

The study also flags how organisations are allocating AI savings. Currently, companies reinvest about 39% of AI cost savings into technology and infrastructure, compared with just 30% directed toward workforce development.

Regional differences are pronounced. North American organisations are least likely to reinvest in people (64%), compared with 84% in EMEA and 89% in APAC.

Encouragingly, four in five respondents agree that organisations that channel productivity gains into workforce development will be more competitive and resilient over the long term.

Rethinking AI success metrics

The findings suggest companies may need to rethink how they measure AI success. Rather than focusing solely on time saved, the report recommends tracking net value by factoring in both efficiency gains and rework costs.

Researchers also advise organisations to update job descriptions to clarify where AI should assist and where human judgment remains essential — a step seen as critical to closing the AI productivity gap.

As AI adoption accelerates, the message from employees is clear: without better training, role redesign, and quality controls, the promise of AI-driven productivity may continue to fall short of expectations.