Furman Says He’d Choose AI Over 20 Economists for Much of CEA Analysis

Jason Furman said a CEA he led today would use AI for a larger share of its analytical work than a hypothetical team of 20 economists, while keeping people in meetings, press and congressional work.

Furman, a professor of economic policy at Harvard, framed the comparison as hypothetical. The 20-person team would include economists with current and future Ph.D.s; the figure does not describe the CEA’s headcount or a hiring plan. “I would somewhat regretfully and sadly choose the AI systems,” he said.

Furman said the comparison concerned a larger share of analytical tasks, rather than every part of the agency’s work. He said AI’s reliability depended on proper fact-checking and that it remained less creative than the best human researchers. The hypothetical did not describe an announced CEA plan to replace economists with AI.

The CEA advises the president on domestic and international economic policy. The agency is part of the Executive Office of the President and was established by Congress in 1946. Furman chaired the council from Aug. 4, 2013, to Jan. 20, 2017.

Furman’s Econvitals project went live Oct. 23, 2026. The free resource tracks the macroeconomy and offers tools for following FOMC speeches and fiscal forecasts. It generates economic briefings using Claude and Briefwright, an engine Furman built in September 2026.

The briefings pass through two automated checks. One compares figures with their sources; the other checks claims against an evidence ledger. No person reviews the briefings before delivery or posting, and they may contain mistakes. The workflow is an example of AI-generated economic analysis, but it does not establish how AI would perform across the CEA’s policy work.

Former CEA Chair Jared Bernstein emphasized the role of context in the agency’s work. “CEA work, at least in my experience, is highly bespoke,” Bernstein said.