- Employees using AI tools report spending significant time refining generated output, according to new research.
- Nearly half of respondents have had to make substantial corrections to AI-produced material.
- Over a third say verifying AI work takes longer than completing the task manually.
Is artificial intelligence consuming more time than it saves? That appears to be the prevailing sentiment among workplace AI users, according to a new study examining the technology’s impact on productivity. The short answer: not as well as hoped.
The Use.AI study surveyed over 8,000 adults who use AI in professional settings across the UK, European Union, US, Canada, Australia, and Latin America. While the technology made it easier to generate documents, code, and memos—with 62% of respondents praising that ease of creation—it also demanded heavy revision. Almost 48% reported that AI creates more problems than it solves, frequently producing material requiring significant correction.
The Editing Burden
The research highlights a “fire and forget” approach to AI-generated content. A concerning 34% of respondents admitted to forwarding AI-assisted work without reviewing it, a practice that leads to errors when the output is unsuitable or incoherent. However, those who do verify the output encounter a different bottleneck: 48% received AI-assisted work requiring corrections, and 43% were forced to completely redo unsatisfactory material.
This verification loop erodes efficiency. The survey found that 39% of workers spend more time checking AI output than they would creating the content from scratch, while 41% noted a sharp increase in the review and verification workload since adopting AI tools.
Is AI Delivering a Productivity Boost?
These findings raise critical questions for organizations that have invested heavily in AI tools, implementation policies, and “AI-ready” hiring strategies over recent years. Rather than accelerating workflows, the research suggests AI often introduces a new layer of oversight. More management attention is required to ensure accuracy, meaning hours are diverted to resolving errors and quality issues—effectively negating the anticipated productivity gains.
Perhaps the most revealing statistic: 58% of respondents stated their organizations have no rules for verifying AI-assisted work. This governance vacuum likely explains the proliferation of unreviewed, low-quality “AI slop” distributed after a single prompt. Without structured validation processes, the efficiency promise of generative AI remains largely unrealized.
