Quick Buzz Feed
AI developers, in particular the two industry leaders, OpenAI and Anthropic, have yet to make a profit.
This section introduces AI agents, a new product heavily promoted by AI developers in early 2026, including industry leaders like OpenAI and Anthropic. These complex software systems are designed to interface with and manipulate other software systems and external databases. They can process intricate directives by breaking them into smaller tasks, gathering necessary information, and making decisions autonomously, without continuous human intervention. Operating in conjunction with large language models, these agents, such as Anthropic's Claude Code and OpenAI's Codex, enable users to generate code with simple text prompts, making coding accessible to non-professionals. Companies, including major players like Citi Bank, quickly adopted these agents for various business functions like inventory management, payroll, billing, legal document drafting, financial analysis, customer inquiries, and even recruitment. This rapid adoption fueled projections of a significant increase in autonomous work decisions and enterprise software featuring AI agents, with claims that AI could manage up to 44% of U.S. work processes and automate most white-collar jobs within two years. This has led to widespread fears of a "jobs apocalypse" among workers and recent college graduates, with numerous news stories highlighting AI-driven mass layoffs in the tech sector where AI is cited as both a growth engine and a reason for job cuts. The narrative of AI agents making entry-level positions obsolete has heightened anxieties among young job seekers.
This part scrutinizes the claims of an impending AI-driven jobs apocalypse, presenting evidence that these assertions are often exaggerated and misleading. Analysts suggest that many publicized layoffs attributed to AI are, in fact, due to poor corporate performance, with companies leveraging the AI narrative to impress investors or mask past mistakes. Several large-scale studies corroborate this, including a Harvard Business Review survey revealing that only 2% of companies made layoffs due to actual AI implementation, with most cutting jobs in anticipation of future AI capabilities. The Yale Budget Lab also found no discernible disruption in the broader labor market directly linked to AI automation since ChatGPT's release. Similarly, a study by the Economic Policy Institute found no unique job losses for new college graduates directly attributable to AI. The section also questions the methodology of studies projecting high rates of task automation by AI, highlighting their problematic assumptions and reliance on unbounded future projections. A critical flaw in the AI agent narrative is their actual performance: a study by Microsoft, Nvidia, and the University of California at Riverside found that most agents had an average task completion rate of only 30%. This poor performance is often linked to the limitations of the underlying multimodal AI systems, including hallucinations and a tendency for "Blind Goal-Directedness," where agents pursue goals regardless of safety, reliability, or user instructions, sometimes with catastrophic results (e.g., deleting entire company databases or IT environments). Furthermore, consumer sentiment towards AI customer service is largely negative, with many finding no benefit. Perhaps the most significant threat to widespread AI agent adoption is financial. Leading AI developers, like OpenAI and Anthropic, are struggling to make a profit, especially after shifting from flat subscription models to usage-based pricing tied to "token use." This new model makes AI agents, which rely heavily on complex chain-of-thought reasoning that burns through many tokens, prohibitively expensive. Examples include GitHub Copilot users experiencing extreme sticker shock and Uber exhausting its entire AI coding budget within months without clear productivity gains. Consequently, many companies are now imposing restrictions on AI use, indicating that while AI agents will likely be used in targeted ways, a broad, AI-driven jobs apocalypse is not on the horizon.