A little over a decade ago, the author led the controls design for a digital-control system intended for a U.S. nuclear plant. The design included manual steps within automated sequences to ensure operators maintained their skills. The project was ultimately shelved due to external factors unrelated to engineering, but the design philosophy remains relevant today as discussions arise about the impact of AI on human expertise.
Research from Harvard University indicates that after the adoption of generative AI, junior employment in companies fell by approximately 9 percent within six quarters compared to non-adopters, while senior employment continued to grow. A Stanford analysis supports this, showing that younger workers in AI-exposed jobs lost ground after late 2022, while more experienced workers maintained their positions. The reasons for this trend are debated, with some attributing it to remote work rather than AI.
The core issue is that expertise is developed through experience, including failures and challenges that AI can bypass. If junior engineers are shielded from essential learning experiences, they may lack the necessary skills to handle critical situations when automation fails. This phenomenon is known as the automation paradox, where increased automation reduces human practice, leaving operators unprepared for unexpected challenges.
The aviation industry has learned from similar experiences, such as the crash of Air France flight 447 in 2009, where pilots struggled to respond effectively due to a lack of recent manual flying experience. In response, the FAA issued guidelines promoting manual flight operations to counteract skill decay.
The author proposes a concept called the "manual gate," which involves intentionally incorporating manual tasks into workflows to preserve essential skills. For instance, in a software team using AI for coding, a manual gate could require junior engineers to debug without AI assistance before consulting the model for solutions. This method ensures that engineers maintain their problem-solving skills and are prepared for situations where automation may fail.
While implementing manual gates may seem less efficient in the short term, it serves as a form of insurance against future skill loss. Organizations that prioritize preserving expertise may need to adopt this approach, even if it conflicts with short-term financial pressures. The author concludes that deliberate inefficiency can be a strategic investment in long-term capability, ensuring that human operators remain competent and ready to take control when necessary.