On the side of waiting, 56% of US adults are extremely or very concerned about AI-driven job loss, according to a Pew Research Center survey of 5,410 US adults fielded in August 2024. In the same survey 64% believe AI will lead to fewer jobs over the next 20 years. Only 23% think it will have a positive impact on how people do their jobs. No direct regret survey exists for proactive AI retraining decisions. The rates here are therefore proxies built from forward-looking concern and aspiration, not retrospective measurements.
The risk is concentrated. Approximately 19% of the US workforce works in the most AI-exposed roles, by Pew’s 2023 occupational analysis, and these workers face the highest structural risk. Their jobs tend to pay more ($33/hour on average, against $20/hour in the least-exposed jobs), so for them the displacement stakes of waiting are significant.
Waiting is a bet that the disruption fails to materialise, and also a bet on timing, because retraining carries its own risk. The AI tools and skills most valuable today, such as prompt engineering, workflow automation on a particular platform or specific coding assistants, evolve faster than most training curricula. A worker who retrains now for one AI application may find parts of that skill set superseded within a few product cycles. Given the pace of AI capability change this is a plausible obsolescence risk, but no study has yet tracked how often it actually happens to early retrainers. Early retrainers may find that a one-time credential is not enough, and that the real choice is between starting continuous learning earlier or later.
Most of them would also be doing it alone. Only 36% of organisations operate as career development champions with structured learning programs (LinkedIn 2025 Workplace Learning Report). The majority of workers who retrain therefore do so without employer support.







