Why AI May Replace Researchers Before Engineers, According to an OpenAI Insider

A surprising claim about the future of jobs inside artificial intelligence labs is sparking debate across the tech world. According to comments attributed to an OpenAI employee, research roles — often considered among the most intellectually secure — may be automated by AI before engineering or sales jobs.

The claim was shared publicly by Yuchen Jin, co-founder and CTO of Hyperbolic Labs, who said an OpenAI researcher told him that researchers could be the first roles replaced by AI within the company. Jin posted on X that, internally, the belief is researchers would be automated first, followed by infrastructure engineers, while sales teams would remain human-driven the longest. Although the individual who made the comment was not named, the statement quickly drew attention for challenging long-held assumptions about which jobs are safest from automation.

At first glance, the idea seems counterintuitive. Engineers are often seen as more vulnerable to automation than researchers. However, Jin explained why research roles may be more exposed than expected. He argued that much of day-to-day research work involves generating ideas, designing experiments, testing variations, and analysing results — tasks that modern AI systems are increasingly capable of doing at remarkable speed and scale.

Jin pointed out that most research follows established patterns, making it easier for AI to replicate. Advanced models can already generate hypotheses, run simulations, and process data far faster than humans. He did, however, draw an important distinction: elite researchers who consistently push the boundaries of knowledge may still be difficult to replace. Their ability to frame entirely new problems remains a uniquely human strength.

Infrastructure engineers, on the other hand, occupy a more resilient position. Jin described AI infrastructure as vast, chaotic, and riddled with edge cases. These systems often rely on highly customised codebases that do not exist in public datasets used to train AI models. Errors in this environment can be extremely costly, making human oversight critical. While AI can write code, managing, debugging, and maintaining large-scale infrastructure remains a complex challenge.

Sales roles appear to be the safest — at least for now. Jin characterised sales as deeply rooted in human psychology, trust, relationships, incentives, and emotion. These are areas where AI continues to struggle with consistency and nuance. He even joked that sales is “the final boss” for AI, suggesting it may be one of the last professions where humans clearly outperform machines.

The comments resonated with other professionals online. Software engineer Sergey Nikiforov admitted he previously assumed infrastructure roles would be automated first but said Jin’s reasoning changed his perspective. Jin responded by emphasising how disorganised and complex AI lab infrastructure really is, noting that teams building systems from scratch may have an advantage.

This discussion comes amid significant churn at OpenAI. Over the summer of 2025, the company reportedly lost several prominent researchers to Meta’s heavily funded Superintelligence Lab. High-profile departures included Jason Wei, Zhiqing Sun, Hyung Won Chung, and Shengjia Zhao, a co-creator of ChatGPT and GPT-4. Other notable exits followed earlier leadership changes, leaving CEO Sam Altman as one of the few remaining original founders.

The broader trend of AI replacing human roles is already visible across Big Tech. Microsoft, Amazon, IBM, and Salesforce have all acknowledged workforce reductions linked to AI-driven efficiency. While companies argue AI frees up resources for growth, the OpenAI comments add a new twist — suggesting that even the most intellectual roles may not be immune.