48% of workers report needing substantial corrections to AI-generated workplace content
A global survey reveals that nearly half of AI users must significantly revise outputs, with many spending more time verifying than creating.
Key points
- % of respondents say AI-assisted work requires substantial correction.
- % report checking AI output takes longer than doing the task manually.
- % forward AI-produced work without verifying its contents first.
A Use.AI study of 8,316 adults across the US, UK, EU, Canada, Australia, and Latin America indicates that workplace AI often creates more problems than it solves. Almost half of those surveyed stated they had to make substantial corrections to material produced or assisted by artificial intelligence tools.
While 62% found AI made creating documents, memos, and code easier, the subsequent verification process is burdensome. According to TechRadar, 43% of professionals were forced to redo unsuitable AI work entirely. This rework consumes significant operational time, undermining the initial productivity gains promised by these technologies.
The verification burden
The research highlights a critical inefficiency in current AI workflows. Thirty-nine percent of respondents said checking AI efforts took longer than creating their own version from scratch. Additionally, 41% noted that reviewing and verifying AI-produced material has become a much larger part of their productivity process since adopting these tools.
Risks of unchecked output
A concerning 34% of respondents admitted to forwarding AI-produced or assisted work without checking the contents. This practice risks distributing unsuitable or unreadable material to clients or colleagues. The "fire and forget" approach to AI generation is proving ineffective, as errors and quality issues require immediate resolution after the fact.
Defender's checklist
Organizations should monitor the ratio of time spent on creation versus verification for AI-assisted tasks. Leaders need to establish clear protocols for reviewing AI output before it leaves the immediate team. Tracking instances where work had to be redone can help quantify the true cost of AI adoption and identify training gaps.
Operational impact
Teams usually need to reassess their quality assurance workflows when integrating generative tools into daily operations. Managers should check whether staff are spending more time correcting errors than performing original work. It is essential to inform department heads about the potential for increased review cycles. Organizations must determine if current staffing levels can absorb the added verification load without impacting service delivery or project timelines.
What to do and how to stay safe: Use.AI
- Establish a mandatory review step for all AI-generated content before it is shared externally or used in critical processes.
- Train staff to recognize common AI errors and hallucinations to reduce the time spent on initial corrections.
- Monitor productivity metrics to ensure that time saved on creation is not lost during the verification and rework phases.
- Update internal policies to clarify that employees remain responsible for the accuracy of AI-assisted outputs they submit.
Step-by-step guide: Implementing AI Usage Policies at Work: A Step-by-Step Framework
General security guidance from the InfoSec Ledger newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
How many workers say checking AI takes longer than doing the task?
According to the Use.AI study, 39% of respondents said checking AI output took longer than creating their own version from scratch.
Do most people verify AI work before sending it?
No, 34% of respondents said they forward AI-produced or assisted work without checking the contents first.
What percentage of workers have to redo AI work?
43% of respondents stated they were forced to redo unsuitable AI work, while 48% said it required substantial correction.
