AI productivity and the destination of time

The World Bank's October 2026 South Asia Economic Update, Adopting AI for Growth, supplies a timely reason to ask where productivity gains might lead. Its research compares formal firms' AI adoption and their expectations of future gains. In the surveyed population, 23.4% of Indian firms reported using some form of AI, compared with 42.7% in the United States. World Bank, p.33, Figure 1.16

The study's scope matters. Its AI survey included 1,355 Indian and 401 US formal firms with at least five employees. It re-interviewed firms from the Enterprise Surveys sample and reported weighted estimates. Unpaid household responsibilities sit outside that firm-survey population. World Bank, p.50 and Annex 2.1, p.72

Indian firms' mean expected three-year gain in sales per worker from faster AI development was 3%. The report defines this as the difference between fast- and slow-development scenario expectations, with employment held constant in each scenario. It is not observed growth or recovered time. World Bank, p.55, Figure 2.6; Annex 2.1, p.73

Suppose, as a hypothetical example, that a task becomes quicker. Its consequences could take several forms. A business might produce more within the same hours. A worker might finish earlier, take on additional tasks, or spend time learning a new process. Changes in staffing could affect someone else's income and time. At home, a shorter paid working day might allow rest, or it might make more unpaid responsibilities possible.

Distinguishing these outcomes requires observing working hours, income, household activity and experience alongside the production measure.

The IMF's June 2026 Finance & Development article by Amory Gethin and Emmanuel Saez makes a related measurement boundary explicit. Its account of working hours includes unpaid agricultural work but excludes unpaid home services such as cleaning, cooking and caring for family members. Reading a working-hours measure requires knowing which work it contains. Gethin and Saez, How Much Does the World Work

For South Asia's AI debate, the resulting research question is concrete: as productive activity changes, who gains usable time, whose responsibilities expand, and how do those changes relate to security and well-being? India's earlier survey rounds help establish the distribution we need to understand. Their differences cannot identify an AI effect.

Even if we observed where the hours went, we would still need to ask how they were experienced by the people living them.