There has been a lot of speculation and hand-wringing on whether, and how much, AI is affecting productivity. This article (rather long) analyzes the evidence, and highlights the tension between micro- and macro-trends.
A few things stood out to me:
- An emerging macro shift: upward revision in productivity growth in early 2026 (Brynjolfsson)
- AI as "great equalizer": bigger growth in productivity in less-experience or lower skilled workers. This is a surprise to me; my empirical observation is completely opposite: white collared workers (at least the techies I know) seem to be adopting AI tools more...
- The J-curve effect: general purpose tech usually causes lower productivity first, as companies shift their budget towards tooling/processes, leading to a dip/stagnation
- job-level bottlenecks: jobs that are human-driven (e.g. meetings, reviews, coordination etc) act as bigger bottlenecks to new AI-driven processes
- endogenous adoption and training gaps: AI use today largely driven endogenously (although top management is now forcing AI diktats) but rank-and-file workers are still not properly trained.
- significant gains in specialized fields: notably: software development, etc
- AI agents: still very much in infancy
- Innovation vs Efficiency: speculation that productivity more due to newer products (AI tools) rather than efficiency (which is where the bigger gains are)
There has been a lot of speculation and hand-wringing on whether, and how much, AI is affecting productivity. This article (rather long) analyzes the evidence, and highlights the tension between micro- and macro-trends.
A few things stood out to me:
- An emerging macro shift: upward revision in productivity growth in early 2026 (Brynjolfsson)
- AI as "great equalizer": bigger growth in productivity in less-experience or lower skilled workers. This is a surprise to me; my empirical observation is completely opposite: white collared workers (at least the techies I know) seem to be adopting AI tools more...
- The J-curve effect: general purpose tech usually causes lower productivity first, as companies shift their budget towards tooling/processes, leading to a dip/stagnation
- job-level bottlenecks: jobs that are human-driven (e.g. meetings, reviews, coordination etc) act as bigger bottlenecks to new AI-driven processes
- endogenous adoption and training gaps: AI use today largely driven endogenously (although top management is now forcing AI diktats) but rank-and-file workers are still not properly trained.
- significant gains in specialized fields: notably: software development, etc
- AI agents: still very much in infancy
- Innovation vs Efficiency: speculation that productivity more due to newer products (AI tools) rather than efficiency (which is where the bigger gains are)