43% of pros redo tasks after AI fails, as verifying output often takes longer than manual work

A recent survey indicates that many professionals spend more time fixing artificial intelligence errors than they would have spent completing the tasks themselves. This finding challenges the assumption that automated tools always accelerate workplace productivity.

How it unfolded

• The Use.AI organization surveyed 8,316 adults across multiple regions regarding their workplace experiences with artificial intelligence technologies.
• Respondents reported that while AI simplified the initial creation of documents and code, the subsequent verification process introduced significant delays and quality issues.
• The data revealed that nearly half of the participants found the technology created more problems than it solved, despite initial enthusiasm for its ease of use.

Who is affected

The study covered adults in the United States, Canada, the United Kingdom, the European Union, Australia, and Latin America who actively use AI tools at work. According to TechRadar, 62% of these users appreciated how much easier AI made generating memos, documents, and code. However, this initial benefit was offset by downstream issues. Forty-eight percent stated they received AI-assisted work that required correction, and 43% said they were forced to redo unsuitable AI work entirely.

What happens next

The research suggests that businesses investing heavily in AI tools and related policies may be facing unintended productivity losses. Instead of serving as a pure advantage, AI can become a problem requiring more oversight to ensure tasks are fulfilled accurately. This means employees are spending additional time resolving errors and quality issues. Thirty-nine percent of respondents said checking the AI’s effort took longer than creating their own version from scratch, and 41% highlighted the increased reviewing and verifying required since AI was introduced.

The background in plain language

Artificial intelligence tools generate content by predicting likely text or code sequences based on vast amounts of training data. These systems do not understand context, intent, or factual accuracy in the way humans do. They can produce plausible-sounding but incorrect or inappropriate material, known as hallucinations. Teams usually need to implement rigorous review processes to catch these errors. Checking whether the output aligns with specific business requirements and factual reality is essential. Without human oversight, AI-generated material may contain subtle errors that compromise quality or compliance.

What to do and how to stay safe: Use.AI

• Establish mandatory review protocols for all AI-generated content before it is shared externally or used in critical workflows.
• Train staff to recognize common AI errors, such as fabricated facts or inappropriate tone, to improve verification speed and accuracy.
• Monitor time spent on reviewing AI output versus manual creation to identify tasks where AI does not provide a net productivity gain.
• Document specific failure modes of AI tools in your environment to create targeted guidelines for when to use or avoid automation.

Step-by-step guide: Injection attacks examples: how prompt leaks break AI systems

General security guidance from the Circuit Observer newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

How many people were included in the survey about AI productivity?

The Use.AI study surveyed 8,316 adults who use AI in the workplace across the US, Canada, UK, EU, Australia, and Latin America.

What percentage of workers forward AI work without checking it?

According to the source material, 34% of respondents were forwarding AI-produced or assisted work without checking the contents.

Do users find AI makes creating documents easier?

Yes, 62% of respondents were enthused by how much easier AI made creating documents, memos, and code, despite the need for later revisions.