# AI skills gap *I’m increasingly more worried about an AI skills gap than an AI unemployment crisis* By [egutierreza](https://paragraph.com/@egutierreza) · 2026-04-22 latam, ai, skills, anthropic, latin, vc --- The AI doom discourse is everywhere. Mass layoffs. White-collar automation. There is one problem with this narrative: the data does not back it up. At least not yet. What the Labor Market Data Actually Shows ----------------------------------------- Despite relentless headlines about AI replacing jobs, aggregate labor market data across most developed economies does not show a wave of AI-driven mass unemployment. Employment is still relatively high. Layoffs in AI-exposed sectors are not dramatically outpacing historical norms. The disruption that everyone is predicting has not shown up in the unemployment statistics. Even Anthropic, whose AI models sit at the center of these conversations, published research showing no statistically significant spike in unemployment in the sectors most exposed to AI. That is a notable finding from a company with every incentive to tell a transformative story about its own technology. What the Anthropic study does flag, and what the broader data confirms, is something more specific and more troubling: younger workers are facing a harder path. Entry-level roles that historically served as on-ramps to career development are getting compressed. The LatAm Picture ----------------- For Latin America, the automation exposure numbers are more moderate than the global conversation suggests. Estimates put somewhere between 13% and 22% of workers in the region at meaningful exposure to automation or augmentation through generative AI. That is not a trivial number, but it is not an existential one either. The distributional pattern, though, is what demands attention. They are younger, more educated, more urban, and disproportionately women. These are the workers who entered the formal economy recently, often in service and knowledge-work roles, and who are now finding that the skills they trained for are exactly the ones that AI tools are most capable of augmenting or replacing. I will not pretend the environment is easy. And I will not go as far as some people like Marc Andreessen in arguing that AI will not replace jobs; there are sectors and roles where it already is. But that is not the most important thing happening here. The Real Risk Is Distributional ------------------------------- The most important question is not whether AI destroys jobs in aggregate. It is who captures the productivity gains. If AI-driven productivity improvements flow primarily to capital owners, to firms in high-income countries with the infrastructure to deploy these tools at scale, and to workers who already have strong technical foundations, then AI accelerates existing inequality. The regions and populations that are already behind fall further behind. If, on the other hand, AI access is genuinely democratized (through affordable tools, local-language models, digital infrastructure investment, and skills programs that reach workers in lower-income markets) then the story looks different. LatAm's large underused talent base and under-digitized service economy actually becomes an asset. The productivity gap between a highly digitized workforce and an underserved one is enormous. Closing that gap through AI access is a real opportunity, not just a policy aspiration. The three levers that determine which version we get are access, skills, and governance. **Access** means affordable compute, reliable connectivity, and AI tools that actually work in Spanish and Portuguese, reflect local business context, and are priced for markets where $20/month subscriptions are a meaningful expense. **Skills** means investing in AI literacy not just for developers, but for the workers in services, logistics, healthcare, and education who stand to benefit most from augmentation, if they know how to use the tools. **Governance** means regulatory frameworks that create clarity and trust, without locking out smaller players or creating compliance burdens that only large institutions can manage. LatAm has a real shot at the upside here. The talent pool is large, young, and educated relative to income level. The service economy is underpenetrated by digital tools. The willingness to adopt new technology, when it is designed for the market, is proven. But that upside requires deliberate effort on all three levers simultaneously. What Are You Doing to Close the Gap? ------------------------------------ If you are building, hiring, or teaching in Latin America right now, I am genuinely curious what you are seeing. Are AI tools reaching workers who are not already at the digital frontier? Are skills programs keeping pace with what employers actually need? Where are the gaps widest? This is the distribution challenge of the AI era. I do not think the answer is pessimism, and I do not think it is naive optimism either. It is serious, intentional work. --- *Originally published on [egutierreza](https://paragraph.com/@egutierreza/ai-skills-gap)*