On One's Own Account

A global consultancy has calculated how much of Latin America's work artificial intelligence could do. Bolivia made it into the count and fell out of the analysis, because the model reads job postings. And here, half the country is its own employer. Nobody posts a job ad for themselves.

sala.red · Cochabamba

What was published

On July 9 the McKinsey Global Institute released 67 pages on how artificial intelligence will reshape work in Latin America. It is the third instalment of a series that has already covered the United States and Europe.

The headline figure: 57% of the hours currently worked in the region could be done by a machine or a program using technology that already exists. Roughly 39 points come from agents, programs that work on a computer, and 18 from robots, machines that work with a body. Estimated value: around 450 billion dollars a year by 2030, of which Mexico and Brazil take three quarters.

Bolivia is in that calculation, drawn from the INE's Continuous Employment Survey, third quarter of 2025. It sits alongside Ecuador and Honduras: the countries where the largest share of people work physically. About 40% of its automatable work is physical, which means robots, and robots cost money.

Why we weren't measured

The useful part of the study, the part that says which skills will change and which will gain value, was built from five countries: Argentina, Brazil, Chile, Colombia and Mexico. Those five get dashboards, indices, time series. Bolivia gets none of it.

The report explains that it chose the markets where data were sufficient. Those data come from job postings published online.

A job posting is generated by an employer looking for somebody else. Where half the population works on its own account, there is nobody to post anything. That silence measures the structure of the country, not a deficiency in it.

There is a second reason, and it runs deeper. Postings appear when someone leaves and has to be replaced: they measure turnover. A market stall in La Cancha held for thirty years and handed down to a daughter generates zero postings across three decades of continuous work. The model reads that permanence as missing data.

The statistics that do exist tend to get read one way. They admit at least two.

What a machine can eat

The construction of the model is worth pausing on. McKinsey built it from occupations and wages published by the US Bureau of Labor Statistics, detailed work activities from a database called O*NET, and skills scraped from job postings. That structure was then mapped onto each of the fifteen countries. To generate the crosswalks, McKinsey used OpenAI's GPT-4o, processing each combination as an individual API call.

The property runs in both directions. A labour market that publishes itself online is a market a model can read, learn and replace. Legibility is the precondition for automation, and Bolivian work is largely illegible to a machine.

That coexists with the other thing, unresolved. The Bolivian state built its social protection around the wage contract, which covers 37.5%. That means pensions, health insurance, the year-end bonus, severance. The own-account majority organised its own: guilds, market associations, cooperatives, rotating credit circles, family. It has worked for decades and it absorbs shocks. It does not include a pension.

The law arrives before the data

On October 22, 2025 the Senate approved Bill 178/2024-2025, "Promotion, Management and Use of Artificial Intelligence". It is still waiting in the Chamber of Deputies. Thirty-one articles: they create an AI Regulatory Authority (ARIA), set twelve plurinational principles and expressly prohibit seven things, including mass facial recognition without a judicial order. AGETIC, the state digital agency, gets 90 days to write the regulations and 180 to write the National AI Strategy.

Bolivia will write that strategy without a single domestic time series on what AI is doing to its labour market. The Latin American AI Index 2025, from CENIA and ECLAC, places the country on the bottom rung, among the "explorers", alongside Paraguay, Guatemala, Honduras and Venezuela. That index measures how closely a country resembles infrastructure designed elsewhere.

Guessing at better systems

A model built from here would measure other things. These five exist, they work every day, and they appear in no official statistic:

• Duration. How many years a person sustains their activity. A posting-based market never produces this figure, and in Bolivia it would measure entire careers of thirty and forty years.
• Density of trust. How many transactions are honoured without a signed contract. In La Cancha and the Feria 16 de Julio it runs to millions a week.
• Absorption capacity. How many people the system relocates after a shock. Between 2024 and 2025 real income fell 13.4% and urban unemployment held near 2.3%.
• Transmission. How a trade passes from one generation to the next with no school, no manual and no certificate. It is the largest training mechanism in the country.
• Active reciprocity. Ayni, mink'a and faena move real labour at scale with no money involved.

None of the five fits a model trained on job postings. Building those measurements is unfinished work, and it is work that can only be done from the inside.

The mineral

The report's last figure connects to the other investigation we have been running in this section. Among the hundred most in-demand skills in the region, mineral processing sits near the high end of the automation exposure index. It is one of the capacities set to change most in the coming years.

That technological future runs on lithium. The batteries, the electric cars, grid storage. The largest lithium resources on the planet lie under the Salar de Uyuni: 21 million tonnes, about 20% of everything identified worldwide, per the 2025 USGS figures.

The precise word is resources. A resource is what sits in the ground. A reserve is what current technology can take out. Bolivia holds the first and almost none of the second. Uyuni brine carries a high magnesium content, and that keeps output near 2,000 tonnes a year, less than a tenth of one percent of the world. Chile pulls two hundred times more from a smaller resource base.

Bolivia has documented presence in 47 of the 60 critical minerals on the 2025 USGS list. On McKinsey's value map it appears as one of the small circles. For its work and for its lithium.

The model measures where value is captured. Not where the lithium is, nor who does the work. This country supplies both to the technological future. The model counts neither.