An analysis by Yale Budget Lab found no clear link between AI use and employment, while separate experiments by economists show how the technology is being tested on policy and research tasks.
The lab's Oct. 14 analysis found no connection between its measures of AI use and employment or unemployment. Shifts in the mix of occupations also had not clearly tracked AI's arrival in the workforce. Yale Budget Lab plans to update its analysis regularly.
Former Council of Economic Advisers (CEA) Chairman Jason Furman wrote in a Nov. 1 essay that an AI-supported team of about 10 could handle much of the council's analytical work. During his mid-2013 to early 2017 tenure, the CEA had roughly 35 staff. Furman presented this as his assessment of AI's capacity, not a staffing proposal.
In a separate Paperwright experiment, Furman said a system he built produced a 65-page economics paper about six hours after he supplied a short prompt. He judged the result below the standard for publication in a mid-level journal. Furman says the technology's effect depends on the task: analysis and writing are more exposed to automation than meetings, testimony and policy persuasion. He has also pointed to farmers and accountants as examples of technology's differing effects on labor demand.
Former CEA Chairman Jared Bernstein wrote that “the human skill of reading-the-room is paramount.” Axios reported that former Federal Reserve economist Claudia Sahm asked where future senior leaders would come from if junior roles disappeared. Axios reported that former CEA economist Casey Mulligan said AI tools he built in two federal offices helped increase demand for economists.
Economist Matthew E. Kahn said he submitted a published 2026 paper by Adrien Bilal and Diego R. Känzig to RefineInk, which returned a 28-page critique. The Quarterly Journal of Economics paper estimates that a 1°C increase in warming would cut world GDP by more than 20% in the long run. Its business-as-usual scenario implies a present welfare loss above 30% and a social cost of carbon above $1,200 per ton. Kahn described his own re-estimation as preliminary and said it did not overturn the paper's results.
In Development and Asterisk announced an essay contest on AI's effects in low- and middle-income countries, including labor markets, government and economic convergence. Entries must be 2,000 to 5,000 words. The organizers allow AI for research, translation and feedback, but not to generate essay text beyond translation. The $10,000 contest closes Jan. 1, 2027, at 6:59 a.m. ET (11:59 UTC), equivalent to Dec. 31 at 11:59 p.m. Anywhere on Earth.
