Artificial intelligence could put pressure on prices before it improves productivity, and the eventual effect on inflation will depend partly on who receives the gains, Bank of Italy Governor Fabio Panetta said on Monday.
In a speech on the economic effects of AI, Panetta said investment in computing capacity, energy, specialized labor, and other scarce inputs is already strengthening demand in countries at the technological frontier. Until supply expands, those pressures can raise relative prices and inflation.
Over a longer period, higher productivity could expand an economy’s capacity to produce goods and services, which would usually ease price pressure. Panetta’s argument was that central banks cannot assume that outcome will arrive quickly or affect all households in the same way.
Automation and new tasks could pull demand in different directions
Panetta set out two possible paths. In one, AI would automate tasks previously carried out by workers, reducing labor demand and shifting income toward owners of capital. In the other, it would create new tasks that complement workers’ skills, raise labor demand, and allow employees to share more directly in productivity gains.
Those outcomes would have different consequences for spending. Workers who expect higher earnings may consume more, adding demand before productivity gains are fully visible. Workers facing greater uncertainty about their jobs or wages may instead save more and spend less, allowing the disinflationary effect of higher productivity to emerge sooner.
The governor did not forecast which path will dominate. He said the effects would probably be uneven across workers, firms, and sectors, with larger companies potentially able to adopt the technology faster than smaller rivals.
Productivity estimates will not settle the policy question
For central banks, the issue is broader than estimating how much output AI might add. They also need to assess how investment, wages, profits, household confidence, and consumption respond as the technology spreads.
That argument follows a concern already visible in recent productivity data. Our earlier coverage of US productivity and labor’s share of output showed how stronger output per hour does not automatically mean that compensation receives the same share of the result.
Panetta also said differences in adoption rates could alter trade, capital flows, and exchange rates. In his view, actual movements in activity, demand, and inflation should carry more weight in monetary-policy decisions than uncertain estimates of an economy’s long-run neutral interest rate, the rate thought to be neither stimulating nor restraining growth.
