As school leaders prepare for the upcoming school year in the Philippines, that gap is becoming harder to ignore. In classrooms across the country, some teachers have quietly found a way to reclaim their time. Tasks that once took more than 20 hours a week—especially lesson planning—are now being completed in just a few hours using artificial intelligence. For many, what used to mean late nights and weekend work has been reduced to a fraction of that effort.

And yet, the relief does not last.

“The challenge is no longer the technology—it’s whether institutions are willing and able to respond to it,” said Christian S. Manansala, an AI researcher and education leader who presented the study in Washington, D.C., and serves as Executive Director of Tagpros.

“We didn’t fail to adopt AI—we failed to build systems that allow it to last.”

The finding puts teachers and systems on different timelines: classrooms are already using AI, while institutions are still trying to catch up.

That reality is now backed by a national study conducted during the recently concluded school year across approximately 200 public schools, presented on April 19 at the International Academic Forum (IAFOR) conference in Washington, D.C., which drew participants from nearly 75 countries. It is the first study from the Philippines presented at IAFOR on this topic.

Conducted in collaboration with the Department of Education – National Capital Region (DepEd–NCR) and the Philippine Normal University, with support from Globe Telecom, Wipro, and Page One Group, the findings point to a simple but uncomfortable truth: the technology works, but the system around it does not.

Teachers in the study reported spending up to 35–40 hours a week on non-instructional work. Where AI was used effectively, that burden dropped sharply. Lesson planning time fell from 22.7 hours to as low as 3.38 hours per week. In some cases, what once took a full day of preparation could be completed in just a few hours.

While the study focuses on teacher workload and efficiency, the findings raise broader questions about how systems support and sustain AI use at scale.

At a time when the Philippines continues to struggle with low mathematics and reading scores in recent PISA results, the findings raise a harder question: if solutions already exist in classrooms, why aren’t they translating into system-wide gains?

The study utilized GabAI, a platform developed by Tagpros, an AI startup incubated under UPSCALE, a joint program of the Department of Science and Technology (DOST) and the University of the Philippines Diliman, enabling both teacher support and a clearer view of how AI is actually being used across classrooms.

Across schools, the pattern was consistent. Teachers were not resisting AI—they were already using it. But without structured guidance, consistent training, and institutional alignment, the benefits appeared—and then faded.

“As an educator for over two decades, I’ve seen how much time teachers spend outside the classroom just trying to keep up,” Manansala added.

“What’s striking is that we now have tools that can significantly reduce that burden—but without system support, those gains remain isolated.”

This gap is no longer theoretical. As the next school year approaches, schools are preparing for the same pressures—heavy workloads, limited preparation time, and uneven instructional support—despite evidence that those pressures can be reduced.

For school leaders, the challenge is immediate: how to support something that is already happening but not fully understood.

For policymakers, the implications are sharper. When adoption moves faster than institutional frameworks, policy is left reacting to practices it cannot fully see.

“Policy cannot guide what it cannot see,” Manansala said.

“Use is already happening. The real question is whether systems are prepared to respond—or continue reacting too late.”

When systems cannot see how AI is being used, they are also less equipped to detect how it may be misused—from everyday classroom practices to emerging risks such as deepfakes.

Earlier this year, these findings were shared with officials from the Embassy of the Philippines in Washington, D.C., including Deputy Chief of Mission Felipe F. Cariño III and Minister and Consul Hans Mohaimin L. Siriban, who expressed support for continued research and engagement. The discussion underscored a broader reality: AI use in education is moving faster than the systems designed to guide it.

The same questions are now emerging beyond the Philippines. In conversations connected to the Virginia Department of Education (VDOE), the Virginia School Boards Association (VSBA), and the Metropolitan Washington Council of Governments (MWCOG), attention is shifting away from access alone toward visibility, governance, and institutional readiness.

For corporate social responsibility (CSR) and funding institutions, the study highlights a different concern. Much of today’s investment in AI for education focuses on access—tools, platforms, and exposure. But access alone does not sustain impact. Without support structures and visibility, gains fade, outcomes are difficult to measure, and scaling remains uncertain.

“We are investing in AI access—but not in the systems that make its impact last,” Manansala said.

What emerges is a shift in the problem itself.

The bottleneck is no longer technology—it is institutional capacity.

Looking ahead, Manansala emphasized the broader national implications of the work.

“We intend to share these findings with the Philippine Legislature and the Office of the President to help inform national AI policy,” he said.

“If supported at scale—through both government and the corporate sector—the Philippines has a real opportunity to position itself at the forefront of AI in education.”

The debate is no longer whether AI should be used—it’s who is responsible for making its benefits last.