AI’s Remote Work Automation Capabilities Still Limited, Study Shows
As artificial intelligence (AI) technologies rapidly evolve, concerns regarding their impact on the job market and employment opportunities are growing. Simultaneously, there’s a keen interest in understanding the extent to which these systems can automate complex tasks, especially in remote work environments, which have become more prevalent since the COVID-19 pandemic. A recent study aims to shed light on the real capabilities and practical limitations of these technologies.
The Study’s Findings
A new study from the Center for AI Safety indicates that popular systems like Manus, ChatGPT, and GPT-5 can only automate 2.5% of complex remote work projects. This suggests that AI’s impact in this area remains very limited. According to the Pew Research Center, nearly a third of U.S. employees fear that AI will reduce their future opportunities, including those working remotely. However, the study’s results offer some reassurance, revealing that AI tools are still unable to perform most tasks with the same level of professionalism as humans.
Methodology and Systems Tested
Researchers tested six systems: Manus, Grok 4, Sonet 4.5, GPT-5, ChatGPT Agent, and Gemini 2.5 Pro. The projects involved product design, data analysis, animation, architectural model development, and scientific paper formatting. The estimated value of these projects exceeded $140,000, representing approximately 6,000 human working hours.
Performance Analysis
The results showed that Manus performed best, with an automation rate of 2.5%, followed by Grok and Sonet at 2.1%. GPT-5 and ChatGPT Agent achieved 1.7% and 1.3%, respectively, while Gemini 2.5 Pro recorded the lowest performance at only 0.8%. The report noted that current AI systems often fail to complete projects to the required professional standard.
Reasons for Failure
The primary reasons for failure included poor quality in approximately 45.6% of outputs and incomplete tasks in 35.7% of cases, such as producing very short videos or inconsistent 3D models.
Future Outlook
Despite these limitations, researchers believe that the performance of these models is gradually improving. The progress in executing complex tasks is becoming measurable, indicating that complete remote work automation remains a long-term but achievable goal.
Source: Annahar
About the Author
