Three banks, three methods, one answer

The Bank for International Settlements (BIS), the Basel institution that acts as a bank for central banks, published Bulletin 130 on 28 July finding that countries scoring high on its AI Preparedness Index (AIPI) saw unemployment rise by 0.75 percentage points on average between 2023 and 2025, while low-scoring countries stayed flat. ¹ The World Bank followed on 4 August with its World Development Report 2026, putting 14.2% of high-income jobs at risk from generative AI automation against 4.5% in low- and middle-income countries. ² The International Labour Organization (ILO), the United Nations agency that sets labour standards, published its youth employment report on 11 August carrying the first headcount scenario in that series: 6.1% of jobs held by 15 to 29 year olds sit in its most exposed occupational categories, and losing a tenth of them would push 5.6 million young people into unemployment, a forced job change, or out of the labour force altogether. ³

None of the three shares a method with the others. Basel ranks national readiness and watches unemployment move. Washington scores occupations for automatability and weights them by national employment. Geneva builds an exposure index with Poland's national research institute and applies it to youth cohorts. Convergence across designs that disagree about almost everything else carries more weight than any one study alone, and it is what the 200 economists who signed Stanford's July call for urgent action had asked somebody to produce .

Footnote 2 of the BIS bulletin then undercuts its own headline, recording that the AIPI is strongly correlated with employment shares in AI-exposed, cognitively intensive sectors. A country scores high partly because it already holds a large number of the jobs at issue, so the index encodes some of the exposure it is being used to predict. Rich economies that score well also ran different monetary policy, different immigration policy and a different post-2023 hiring cycle. The bulletin claims no causation, calls displacement so far limited, names only call centres and business centres as categories showing early signs, and says the productivity payoff remains uncertain and uneven.

One finding inside the same bulletin points forward rather than back. Text analysis of company earnings calls shows nearly 80% of firms now signalling intent to increase labour substitution, which records managerial appetite rather than realised displacement. That appetite is what none of these three institutions can yet see landing in their data.