Artificial intelligence is already changing companies and professions, but its mass adoption will not occur at the pace of financial expectations nor of the most extreme predictions about the future of work. The lack of infrastructure, chip scarcity, electricity and water consumption, and the need to keep decisions under human responsibility slow its expansion.
The debate matters because AI can raise productivity without automatically becoming a general substitute for people. In Mexico, impact will reach sectors unevenly: some activities will incorporate advanced tools, while others will barely modify their processes. The result will depend less on technological enthusiasm than on the ability of companies and workers to adapt.
Automating tasks does not equal eliminating professions
One of the main warnings about AI relates to employment. Banamex has estimated that 30% of formal jobs in Mexico could be automated, although that figure refers to positions or tasks exposed to technological change, not necessarily to immediate layoffs. The difference is relevant: many occupations combine automatable activities with tasks that require judgment, communication and responsibility.
The experience of previous technological transformations also advises caution. The spread of computers did not eliminate work, but modified skills and processes. Likewise, AI can reduce routine tasks and create demand for training, supervision, analysis and design of solutions, although the transition may incur significant costs for those without access to training.
Cuts attributed to AI do not by themselves explain business decisions. In some cases, companies had expanded their staff during the pandemic and later adjusted size for financial reasons. Portraying every layoff as a direct consequence of automation may exaggerate the current effect of technology.
Infrastructure imposes physical limits
AI models need data centers, chips, electric grids and cooling systems. The International Energy Agency estimates that data center electricity consumption increased by 17% in 2025, while AI-oriented centers grew even faster. The technological expansion, therefore, depends on physical investments that take years, permits and energy availability.
Mexico faces this challenge in regions like Querétaro, where data center projects are growing. The demand for electricity and water has raised concerns because installing new capacity competes with other industrial, urban and agricultural uses. Also, chip shortages and political opposition can delay projects that, in the market, are already presented as part of an imminent transformation.
The environmental problem does not invalidate the technology, but forces measuring its costs. The efficiency of an AI system must be evaluated alongside the energy it consumes, the water used to cool it and the infrastructure needed to keep it operating. Without those calculations, productivity promises may conceal impacts borne by communities and public services.
Unequal and gradual adoption
AI will have deep effects on activities with large volumes of information and repetitive tasks, such as certain financial, legal, logistical and administrative services. In contrast, its ability to transform physical jobs, small businesses or processes that depend on trust and responsibility will remain more limited. A tool can draft, classify or predict, but it does not by itself remove the obligation to be responsible for an error.
In Mexico, business adoption still shows delays compared to other economies. This may delay benefits, but also means there is room to define rules on privacy, training, security and accountability before the technology spreads on a larger scale.
AI will probably transform a significant part of the economy, but the process will be more like an accumulative transition than an instant rupture. Its advantage will not necessarily be in producing with fewer workers, but in allowing people to devote more time to solving problems, making decisions and developing activities that machines still cannot reliably undertake.