The PISA paradox in the AI era: How to scale technology when only 28% master basic math

If you follow the Brazilian technology and digital retail market, you have probably noticed a curious paradox: Brazilians are among the most connected people on the planet, spend hours a day on mobile apps, adopted Pix at lightning speed, and rank among the global top users of generative Artificial Intelligence tools. Even so, when we measure the productivity per worker of the Brazilian economy, it remains stagnant at practically the same level since the mid-1990s.

Why does the massive consumption of technology in Brazil not automatically convert into a productivity leap in companies? The most honest answer lies in the reports of the PISA (Programme for International Student Assessment), coordinated by the OECD.

In PISA surveys, only 28% of Brazilian 15-year-old students reach at least level 2 in mathematics, compared to 65% on average across OECD countries and over 80% in economies like Singapore, Estonia, Japan, and South Korea. In reading, about half of our young people cannot identify the main idea of a moderately long text nor distinguish fact from opinion.

For those leading engineering, operations, or business in Brazil, this data is not a distant statistic from the Ministry of Education: it is the technical description of the workforce entering our companies every year.

1. What the lack of level 2 means on the shop floor

On the OECD scale, level 2 does not require differential calculus, linear algebra, or quantum physics. It is defined as the minimum threshold of functional survival in the modern world.

When 72% of young Brazilians leave school below this level, it means that almost three out of every four new workers face severe barriers to performing elementary analytical tasks:

• Understanding the difference between percentage variation and percentage points in a sales dashboard.
• Reading a trend chart or identifying that an inventory indicator is outside the normal curve.
• Following a logical decision tree with conditions ("if A and B occur, execute C; otherwise, trigger D").
• Comprehending simple proportions of dosage, raw material waste, compound interest, or contribution margin.

Even more concerning for the technology and engineering sector is what happens at the top of the distribution (levels 5 and 6 of PISA, where young people capable of modeling complex problems, abstracting systems, and programming advanced algorithms are found). On the OECD average, about 9% to 11% of students reach these levels of excellence; in Singapore and Taiwan, this share exceeds 25% to 35%. In Brazil, it hovers around 1%.

2. The amplifying effect of Artificial Intelligence

Many executives nurtured the hope that Large Language Models (LLMs) and AI copilots would magically solve the Brazilian educational deficit: "if the employee does not know how to write well or does not know how to program and analyze spreadsheets, AI does it for them".

Practical reality inside companies is showing something much more subtle. Generative Artificial Intelligence works as a cognitive exoskeleton, but it demands two critical human skills: precise problem formulation (context and logic) and critical auditing capacity (verification of hallucinations and mathematical errors).

A software engineer or financial analyst who masters statistics, systems architecture, and critical reading multiplies their output by three or five times using AI agents, because they know how to specify exactly what they want and notice in seconds when the model has generated an inefficient SQL query or flawed financial premise. Conversely, a professional without a logical foundation tends to accept incorrect AI outputs passively, automating errors on an industrial scale.

3. How to lead technology and operations in the real Brazil

While we demand as citizens a deep reform in Brazilian basic education (following the example of what Poland, Estonia, and prominent Brazilian municipalities like Sobral did), business and technology leaders need to adapt their engineering strategy to the country's reality today:

1. Systems architecture with native guardrails: when you build internal software for retail, logistics, health, or customer service operations in Brazil, do not design interfaces that depend on ambiguous interpretation or manual calculations by the frontline operator. Embed mathematical intelligence, limit validation, and error prevention directly into the code and interface flow.
2. Use AI to simplify the interface, not to transfer responsibility: instead of handing a blank prompt to thousands of operators, build deterministic AI-assisted flows where data is already structured, business rules are validated by traditional code, and the system guides the collaborator step by step.
3. Turn your company into a academy of fundamentals: the best engineering and data departments in Brazil have discovered that training the team only on the "trendy tool" has a short return. Setting aside hours to teach logic, data reading, the scientific method of debugging, and business-applied mathematics creates a competitive advantage that no competitor can copy overnight.

Continue deepening: also read Poland's leap: how the education and engineering shock unlocked productivity and AI agents and unit economics: how to measure the real return of automation in operations. Also explore our work in Executive Leadership and Fractional CTO and our Case Studies.

---

References and verifiable readings

• OECD (2023). PISA 2022 Results (Volume I): The State of Learning and Equity in Education. Paris: OECD Publishing.
• INEP / Ministry of Education. National Report of Brazil in PISA.
• Hanushek, E. A., & Woessmann, L. (2015). The Knowledge Capital of Nations: Education and the Economics of Growth. MIT Press.