How long will organizations take to redesign processes, software, incentives, and management around AI?
General-purpose technologies historically take decades to show up in productivity statistics, because value requires organizational reinvention. The J-curve predicts measured productivity dips before it surges. The live counter-argument is self-diffusion: unlike past general-purpose technologies, AI might do part of its own integration work — writing the code, redesigning the processes — compressing the historical lag.
View on the map → · Open in Browse →
The adoption numbers stopped agreeing with each other and started explaining why: Census-linked data show narrow deployment inside adopting firms, a four-country executive survey finds 69% adoption with near-zero realized effects, and the St. Louis Fed showed the gaps are partly survey wording. Usage data adds the mechanism — learning curves govern how fast value arrives.
57% use AI in three or fewer functions, most often Sales and Marketing (52%), Strategy (45%), and IT (41%)
2026 Census supplement: 18% of firms use AI, but deployment within firms remains narrow.
nine-in-ten reporting no impact on employment or productivity
Nearly 6,000 executives in four countries: widespread use, almost no reported productivity or employment effects yet.
Apparent adoption gaps are partly artifacts of survey wording; US firm adoption nearly doubles when the question broadens.
Survey of 2,021 Americans: task replacement for 27% of employed AI users, new tasks for 21%.
High-tenure users succeed ~10% more often and attempt higher-value tasks — learning-by-doing governs diffusion speed.
Transformation runs through decades-long innovation-diffusion-adaptation cycles; reliability, integration, and tacit knowledge sit downstream of capability.