- Generative AI has significantly improved in 2025, becoming capable of automating many high-value tasks and operating independently for longer periods, but its impact on jobs remains controversial.
- OpenAI’s GDPval metric shows rapidly increasing AI capabilities: GPT-4o (May 2024) only matched or exceeded humans in 12% of tasks, Claude Opus 4.1 (August 2025) reached nearly 48%, while GPT-5.2 Pro achieved about 74% in digital work tasks.
- However, this benchmark is biased toward digital knowledge work and does not fully reflect teamwork, corporate context, or the entire labor market.
- 2025 studies show that AI has not yet drastically changed total employment: the proportion of jobs with high AI exposure remains stable, and wages in this group have even increased. The Fed Chair suggests AI is “only a small part” of a weak labor market.
- Nearly 90% of business leaders in the US and UK say AI has not affected headcount or productivity in the past 3 years, but the calm surface hides underlying volatility.
- AI-first companies require proof that AI cannot perform a task before hiring a new person; research from Stanford and Harvard shows that junior-level jobs in easily automated professions (coding, customer service) have significantly declined post-ChatGPT, while senior roles remain stable or have increased.
- The concept of “workslop” has emerged: AI-generated content that looks plausible but lacks depth; 40% of workers say they have identified such content within a month.
- AI reduces the necessary “friction” in thinking and hiring, leading to the spread of superficial documents and a flood of job applications, making it difficult for recruiters to screen quality candidates.
📌 Conclusion: 2025 shows that AI has not drastically changed total employment: the proportion of jobs with high AI exposure remains stable, and wages in this group have even increased. AI-first companies require proof that AI cannot do the job before hiring new staff; junior-level jobs in easily automated professions (such as coding and customer service) have significantly decreased post-ChatGPT, while experienced personnel roles remain stable or have increased. The concept of “workslop” has emerged, referring to AI-generated content that looks plausible but lacks depth. AI reduces thought and hiring friction, leading to the spread of superficial documents and a flood of job applications, making it difficult for recruiters to screen quality candidates.
