Former Obama CEA Chair Says AI Could Replace 20 Economists

Jason Furman, who served as the chair of the Council of Economic Advisers under President Barack Obama, stated that he would use artificial intelligence systems instead of hiring 20 human economists if he were running the agency today.

Furman, currently a professor at Harvard, described this potential shift as a decision made "somewhat regretfully and sadly." He clarified that the role of the chair would remain occupied by a human, but the support staff would change dramatically.

The CEA was established in 1946 with the specific mission to provide the president with objective economic advice on domestic and international policy. Its history includes instances where staff members used their positions to challenge executive ideas.

Kenneth Arrow, a Nobel laureate, frequently cited a memo he wrote as a CEA staffer in the 1960s as the most valuable work of his life. In that memo, he opposed a plan to widen the Panama Canal using an atomic bomb, demonstrating the agency's role as a check on presidential impulses.

Furman initially argued that AI could replace the entire junior staff, including recent college graduates and graduate students. He noted that if his own logic applied to his past work under President Bill Clinton, his position as a junior economist at the CEA would have been eliminated.

Following pushback from former staff, Furman adjusted his position to call for an "AI-intensive CEA with only a skeleton staff." He maintained that he could not justify keeping human jobs where the majority of tasks could be performed by AI.

Claudia Sahm, chief economist at New Century Advisors and a former Federal Reserve economist, argues that the debate misses the point. She contends that the goal should be how AI serves the organization's mission rather than which specific tasks it can replace.

Sahm points out that AI models rarely identify flawed ideas, a function typically performed by human staff. Junior economists, many of whom are on leave from universities or detailed from other government agencies, often have less political pressure to agree with the president's priorities.

Sahm recalls a project at the CEA where the agency struggled to explain a Federal Reserve survey statistic showing that nearly half of Americans could not cover a $400 emergency expense. The team pulled multiple data sets and wrote several memos without finding a clear answer, a creative process she believes AI would have executed faster but not better.

She notes that unresolved questions often lead to future innovations, such as the Sahm rule, a recession indicator that developed from her early work at the Fed and CEA. She argues that replacing juniors with AI could disrupt this pipeline of ideas and quality checks.

Furman acknowledges that AI can speed up data work and help tailor economic reports for broader audiences. However, Sahm warns that using AI to monitor bots or handle simple formatting would be a poor use of talent and taxpayer money.

The agency is expected to adapt its staffing model as AI capabilities grow. Furman believes the current tools make a smaller, AI-reliant team viable, while Sahm advocates for a structure that uses AI to free up time for deeper analysis and quality improvements.