The Digital Press

Researchers at Carnegie Mellon University tested an entire fake software company staffed by AI agents, revealing that these AI workers fail to perform most tasks effectively. Despite being sourced from top AI models, the agents struggled with basic job functions and showed significant limitations.

Why it matters: AI agents currently lack common sense and social skills, making them ineffective for complex workplace tasks.

Stunning stat: The best AI completed only 24% of assigned jobs, with some models finishing as few as 1.7%.

The big picture: Present AI is more like advanced predictive text, not a sentient intelligence capable of independent problem-solving or learning.

Commenters say: Many agree AI isn’t ready to replace human jobs and see it as a tool that amplifies human intent, not a fully autonomous workforce.