Making a transformative new technology available is rarely enough to translate its capabilities into real-world value.
The first phase of adoption typically involves fitting the new technology into existing structures, producing only modest gains. The greater gains come only later, when we rethink and rebuild around what the technology makes possible.
Factory electrification is a classic example of this pattern. Early manufacturers simply replaced their steam engines with electric motors while leaving their factories’ shafts, belts, and layouts largely unchanged. Unlocking electricity’s potential required rethinking the factory itself. The breakthrough came when manufacturers gave individual machines their own motors, freeing factory layouts from the constraints of shafts and belts.
AI in education is still in the first phase of adoption. It is widely available and implemented piecemeal within existing structures, but without the deeper redesign needed to realize its transformative potential.
This is my attempt to imagine what rebuilding education around AI might actually look like.
Guiding principles
I believe it’s critical to be intentional about how we implement AI. To clearly state how we want the future to differ from the present, define what we don’t want AI to do, and identify where we see opportunities.
- Teacher-led. Teachers must remain in charge of their students’ educational experience. AI can give each student personalized support and attention that no teacher can provide alone, but it should do so under the teacher’s direction, extending their reach rather than assuming their authority. No matter how capable AI becomes, the human relationship between teacher and student will remain essential to the often overlooked intangibles of schooling like motivation and social development.
- Deliberate simplicity. Students should not have to navigate an ever-growing collection of devices, apps, and services. Introducing AI must reduce this complexity rather than add to it. This means actively stripping out unnecessary tools and sources of distraction. The learning environment should be simple, focused, and purpose-built for education.
- Durable learning. The goal should not be for students to complete more work with AI, but to become more capable because of it. Short-term measures such as task completion and test scores can be gamed and consequently conceal whether genuine learning has occurred. We must judge success by what students retain and can do independently in the long run.
- Socially grounded. AI-driven personalization must not come at the cost of a more isolated educational experience. Education is shaped not only by the instruction each student receives, but also by relationships and shared experiences. AI should be carefully designed to strengthen these social foundations rather than displace them.
Presenting EduOS
I call my solution EduOS. A learning environment, redesigned around what AI makes possible.
It has three parts. The device is the interface. The tutor is what happens on the device. The classroom is how teachers and the class run once both exist.
The device
No problem is as important to getting this right as the medium in witch the the interaction happens between the studenten and the AI.
96% of OECD students have a computer at school. The research around the impact of use with current digital devices is bleak. 65% of students self-reported being distracted by their own use of digital devices in at least some times. And among teachers in Sweden 85% agreed that digital tools distract students from their learning.