AI Layoff Risks Overstated, but Automation Threatens Certain Jobs

David Sachs, White House Cryptocurrency and Artificial Intelligence Adviser, says the risks of AI-related layoffs are overstated. Microsoft has identified the jobs most susceptible to automation. The cryptocurrency industry is experiencing a noticeable reduction in the number of job vacancies. White House Cryptocurrency and Artificial Intelligence Adviser David Sachs said predictions of mass layoffs due to AI are exaggerated, emphasizing that the technology still requires human oversight to create real business value. Sachs wrote on the social network X (formerly Twitter) that AI serves as a “middle-man” in workflows, while key decisions and management remain with humans. He stressed that even the most advanced AI models need human review and guidance. This statement came shortly after Microsoft Research published a study listing the jobs most at risk of AI impact. The list included journalists, analysts, and technical writers — roles heavily involved in the cryptocurrency space. Microsoft researchers analyzed 200,000 anonymous Bing Copilot chats and found that AI is most often used for data collection, writing, and consulting. The highest risks of automation were observed in reporting and copywriting, with an AI applicability score between 0.38 and 0.39. Market analysts and data scientists were rated slightly lower, with scores of 0.35 to 0.36. The scores reflect the likelihood that AI can effectively perform these tasks compared to human workers. The issue is also relevant for the crypto industry. In July 2025, experts noted a significant decrease in the number of vacancies in the sector. This decline coincided with a report from the US Department of Labor, which revealed that only 73,000 jobs were created nationwide, compared to the expected 100,000 forecasted by Dow Jones. Sachs also cited former Coinbase CTO Balaji Srinivasan, who said that AI is not taking jobs from humans but is instead replacing earlier generations of AI models. “Today's AI is not truly agentic. It can perform more tasks, but it does not act completely on its own,” Srinivasan said. He highlighted that competition is primarily between AI models themselves. For example, Midjourney has displaced Stable Diffusion in image generation, and GPT-4 has replaced GPT-3 in large language models. These developments support Sachs’s view that human control remains key to the effective use of AI technology.