The real ROI of AI for knowledge work: the task-level gains (Noy-Zhang, Brynjolfsson), the jagged frontier (BCG-Harvard), the 2.8% real-world time saving and no earnings effect (Humlum), 95% of enterprise pilots with no P&L return (MIT), and how to capture what is real.
This article is AI slop.
The actual paper is here: https://www.nber.org/papers/w33777