What Latin America's AI Framework Exposes About Everyone Else

Key Takeaways

• The Inter-American Development Bank identifies five enabling conditions for AI in education (devices, connectivity, platforms, teacher competencies, and governance) that must develop in parallel, not sequentially. Most education systems, including wealthy ones, are failing this test.

• AI applications require specific connectivity thresholds most schools do not meet: chatbots need minimum 5 Mbps with latency under 100 milliseconds; AI-driven video or voice tools require 10 Mbps with latency under 50 milliseconds.

• Uruguay's Plan CEIBAL succeeded where others failed because they committed approximately 5% of basic education budget annually to ongoing operations, not just initial rollout. Most systems budget for the pilot and forget about year five.

• The question for ministries, boards, and vendors in 2026 is simple: can you produce evidence that all five conditions exist, or are you just buying tools?

I've been in enough procurement meetings to know the pattern. A vendor walks in with a polished demo. The AI does something impressive. Someone asks about data privacy and gets a reassuring answer about encryption. The contract gets signed. Six months later, the tool is gathering dust because nobody trained the teachers, the WiFi can't handle the load, and the governance policy is still "in development."

This isn't a Global South problem. I've seen it in Alberta. I've seen it in well-funded American districts. The technology arrives before the conditions for using it responsibly are in place. Then everyone acts surprised when it doesn't work.

The Inter-American Development Bank just released a technical note on AI in education for Latin America and the Caribbean, and the uncomfortable part isn't what it says about developing nations. It's what it says about everyone. Their core argument is blunt: AI layered on weak digital foundations will repeat the same edtech cycle we've watched for thirty years. Big promises, pilot programs, mixed results, quiet disappointment, then the next shiny thing. The IDB identifies five enabling conditions that must develop in parallel for AI to actually deliver: devices, meaningful connectivity, digital platforms and content, teacher digital competencies, and governance. Not sequentially. In parallel. Most systems are buying tools first and asking governance questions afterward. The IDB calls this what it is: a setup for failure.

The Five-Condition Framework

The IDB framework deserves attention because it treats AI readiness as operational rather than aspirational. Each of the five conditions has specific, measurable requirements.

Devices must be available and functional across the student population, not just in showcase classrooms. Meaningful connectivity means bandwidth sufficient for AI applications, not just email. The IDB specifies that basic educational connectivity requires 1 Mbps per student during peak usage. But AI tools demand more: chatbots require minimum 5 Mbps with latency no greater than 100 milliseconds, while AI-driven videoconferencing or voice assistants need at least 10 Mbps with maximum latency of 50 milliseconds (IDB, 2025, p. 47). Most schools cannot meet these thresholds. Digital platforms must be curriculum-aligned and interoperable with existing systems. Teacher competencies require ongoing professional development, not a single training session before rollout.

Governance is where most systems collapse. The IDB does not treat governance as a philosophy. They treat it as machinery: privacy and data protection rules that are actually enforced, cybersecurity protocols that exist beyond the IT department's documentation, interoperable education management information systems (EMIS) that allow student data to flow securely between platforms while maintaining audit trails, and sustainable financing that accounts for operational expenditure (OPEX) beyond the initial capital investment (IDB, 2025, p. 54-55).

OPEX is where most systems fail financially. Capital expenditure (CAPEX) covers the initial purchase: devices, infrastructure buildout, launch training. OPEX covers everything afterward: maintenance, software updates, ongoing teacher development, technical support, replacement cycles. Everyone budgets for the pilot. Almost nobody budgets for year three, year five, year ten.

Uruguay's Plan CEIBAL is one of the few long-term success stories in the region precisely because they understood this. CEIBAL allocates approximately 5% of Uruguay's basic education budget annually to maintaining its digital transformation, not as a one-time expense but as a permanent budget line (IDB, 2025, p. 54). That commitment to sustainable financing is why the program still functions after fifteen years while similar initiatives across the region have quietly disappeared.

What This Means for Decision-Makers

For ministries, the question is whether you can produce evidence that all five conditions exist in your system right now. Not plans. Not intentions. Evidence. When the audit comes, whether from regulators, insurers, parents, or press, "we were working on it" does not hold up.

For school boards and districts, the challenge is smaller in scale but identical in structure. Vendors have every incentive to downplay the governance burden. They will tell you the tool is plug-and-play. They will not mention that meaningful connectivity in your rural schools is unreliable, that your teachers have not been trained, or that your data governance policy was last updated in 2019. The question boards must ask vendors and themselves is the same one the IDB poses: are the enabling conditions actually in place, or are we just buying tools?

For vendors, this framework is a competitive opportunity. The vendors who win long-term will not be the ones with the flashiest demos. They will be the ones who can walk into a procurement meeting and demonstrate how their tool fits into an existing governance framework, what documentation they provide, and how they support compliance obligations. Most vendors cannot say that. The ones who can will own the market when regulatory pressure arrives. EU AI Act enforcement began this month. California and Colorado have deadlines this year. Canada just allocated $1.26 billion to AI, and the compliance expectations attached to government money are not optional.

I think about that procurement meeting pattern often. The enthusiasm in the room when the demo works. The confidence that "we'll figure out the details later." The silence six months down the road when nobody wants to admit the tool isn't being used.

The IDB framework is a mirror. It asks whether you have done the unglamorous work that makes technology actually function in schools. For most systems, the honest answer is no. The good news is that the work is not mysterious. It is just work. Devices, connectivity, platforms, teacher training, governance. In parallel. With evidence.

If you need a practical system for turning "responsible AI" intent into defensible documentation, that is exactly what my Stop-Gap AI Compliance Guide is built for. Forms, checklists, role-based handbooks, and a year-long compliance calendar that works with existing resources.

And if you want an external assessment of where your organization actually stands, my Assurance Advisory is a fixed-scope engagement that produces documentation demonstrating due diligence. The kind of documentation that matters if a data breach, bias complaint, or parental concern ever becomes a legal question.

Ryan James Purdy is the author of the Stop-Gap AI Policy Guide series and founder of Purdy House Publishing & AI Consulting. With nearly 30 years of experience in education across North America, Europe, and Asia, he helps school boards, ministries, and vendors build AI governance systems that meet legal obligations and survive audits. Contact: jamespurdy624@gmail.com | LinkedIn: linkedin.com/in/purdyhouse

References

Inter-American Development Bank. (2025). AI and Education: Building the Future Through Digital Transformation. IDB Education Division Technical Note. April 2025.