Most Workers Use AI for Tasks Beyond Their Skill Level, Survey Finds

Most Workers Use AI for Tasks Beyond Their Skill Level, Survey Finds

A significant majority of office workers now rely on artificial intelligence to complete tasks that exceed their independent capabilities, according to new research that raises fundamental questions about how organizations evaluate employee performance and readiness for promotion.

A survey of 9,684 working adults across the United States, United Kingdom, Canada, the European Union, and Latin America found that 64% have used AI to complete work they could not have done alone. More than half—52%—said the technology makes them appear more experienced than they actually are, while 43% reported using it to handle responsibilities they don't yet feel qualified for.

The research, conducted by Use.AI, a platform that provides access to multiple large language models, reveals a growing disconnect between delivered results and underlying skill. Thirty-five percent of respondents admitted they would struggle to perform parts of their current job without AI assistance, and a quarter worry their employer overestimates their capabilities.

The disclosure gap

Managers often remain unaware of how much AI contributes to their employees' output. Thirty-nine percent of workers said they have submitted AI-assisted work without mentioning it, while 30% have accepted praise for work the technology substantially produced. Nearly one in five reported that AI-assisted work contributed to a promotion.

Most workplaces lack policies requiring disclosure of AI use, meaning performance evaluations increasingly judge results without understanding how they were produced.

Ihor Herasymov, co-founder and CEO of Use.AI, told Euronews that blanket disclosure requirements would prove impractical as AI becomes embedded in everyday software. Instead, he advocates for a materiality threshold: employees should disclose when AI generates a significant part of an analysis, recommendation, presentation, or other consequential output.

Why it matters

The rise of AI-assisted work fundamentally changes what finished output reveals about employee capability. Organizations making promotion decisions now face a challenge: polished deliverables no longer reliably indicate whether someone can explain reasoning, detect flaws in AI-generated answers, or make sound judgments when the technology fails. As AI autonomy accelerates—with capabilities reportedly increasing 1,400% year-over-year between early 2025 and early 2026—the gap between what employees can produce and what they genuinely understand will only widen, forcing companies to rethink performance assessment entirely.

Rethinking performance assessment

Herasymov acknowledged that finished work has become "a less complete measure of capability." He emphasized that employers need to evaluate whether employees can explain reasoning, detect errors in AI output, and make decisions with incomplete information.

"Problem framing is particularly important," Herasymov stated. "Can someone define the right question, challenge an assumption and explain why one course of action is better than another? That is much harder to infer from a polished final output alone."

Few organizations have established clear frameworks for assessing performance in the AI era. The distinction between what someone can produce with AI and what they actually understand will become increasingly critical in promotion decisions, according to Herasymov.

Tool versus dependency

Asked whether the finding that 35% of workers would struggle without AI concerns a company that sells AI access, Herasymov did not dismiss the issue. He framed it as a question of augmentation versus dependency: AI can extend capabilities and increase speed, but if those gains come with weaker independent judgment, that represents a trade-off employers and technology companies should acknowledge.

The challenge intensifies as AI capabilities advance rapidly. According to data from METR, a US non-profit evaluating AI's societal impact, the time required for AI autonomy to double has compressed from eight months to 4.7 months. Downloads of AI tools surged from 15,000 to 11.8 million—a 780-fold increase—while publicly available Model Context Protocol tools, which enable AI assistants to execute tasks directly rather than just describe them, grew 35-fold to approximately 177,000.

As more work happens without direct human oversight at each step, the finished product reveals progressively less about the person who produced it.

These findings were first reported by Euronews based on research published by Use.AI.