Between 2% and 5% of jobs in Latin America at risk of full automation: the figure calls for context more than alarm

By Publimetro México · November 8, 2026.

Federal deputy Miguel Ángel Guevara Rodríguez, of the PAN, is calling for funds to be allocated to training workers, professionals and young people, and to job-retraining programs for those performing tasks most at risk of automation. The proposal appears in the report on his participation in the Parliamentary Assembly of the Council of Europe (Strasbourg, September 28 to October 2, 2026), published in the Gaceta Parlamentaria, according to Publimetro.

The piece draws on estimates from the World Bank and the International Labour Organization (ILO). According to these organizations, between 2% and 5% of jobs in Latin America and the Caribbean face a risk of full automation by generative AI. Between 8% and 14%, by contrast, could see their productivity improve with these tools.

There is a third key figure: some 17 million jobs in the region that could benefit from AI face connectivity and digital infrastructure constraints. Publimetro stresses that these are regional estimates and not a projection of layoffs in Mexico.

The deputy also argues that when a decision affects a person's employment, safety or rights, final responsibility should rest with a human. As an example of use, he cites a parliamentary adviser who reviews documents in minutes instead of hours.

Our take: the figure that matters is not the 2-5%, but the comparison between the three numbers. Jobs with the potential to gain productivity (8-14%) far outnumber those facing full automation, by roughly three to four times. And the bottleneck, according to the study itself, is not just the technology but access to it. The short-term risk is therefore one of inequality: that those who already have connectivity, tools and training capture the gains while the rest fall behind.

That fits a thesis we hold: AI does not "take away work," it changes which work matters. But that transition is neither automatic nor fair by default. A portion of workers, especially in administrative and repetitive tasks, will face a hard transition. Training is a reasonable response, though insufficient if not accompanied by connectivity and infrastructure. Training someone in AI without reliable internet is a stopgap.

It is also worth being cautious about the scope. This is a proposal in a legislative report, not a law or an approved budget. And the 2-5% measures full automation, not the partial erosion of tasks, which can affect far more jobs without eliminating them.

In the long run, the underlying bet is optimistic: if productivity is well distributed, the abundance AI generates can free people to work on what they are passionate about. To get there, the region first needs to solve the basics: access, training, and ensuring that critical decisions still have a human in charge.