School dropout is rarely sudden. Too often, the risks children face go unnoticed until it’s too late – challenges at home, struggles in the classroom, learning difficulties, limited resources and a myriad of other challenges push children from the learning opportunities that schools should provide. All too often, education systems tend to be reactive and fragmented. 

To address these issues, UNICEF, in partnership with Malaysia’s Ministry of Education, developed a national AI-powered Early Warning System (known as SiPKPM) to predict and prevent school dropout among 5 million students. 

Launched during the ASEAN Out of School Children and Youth Seminar in 2025, this new platform is gaining regional recognition for its innovative, evidence-based solution that goes beyond digital as usual by transforming the way the country identifies and supports children at risk of dropping out. 

The system analyses each student across seven indicators: attendance, academic achievement, disciplinary record, disability or learning support needs, household income, distance from school, and parents’ marital status. Together, these factors help identify which students may be most at risk, sending automatic alerts to counselors, teachers and district officials, so they can step in before it’s too late.

In 2025 alone, the AI system flagged students at risk of dropout, prompting targeted interventions that enabled more than 9,000 students – of whom 3,287 were girls – to return to school.

The indicators themselves were not chosen arbitrarily. They were drawn from an earlier out of school children study by UNICEF and the Ministry of Education, grounding the AI model in evidence based on how and why children in Malaysia drop out of school.

Built for educators, by educators

What distinguishes the SiPKPM early warning system from standalone tech experiments is how it puts teachers and education professionals first by ensuring they have the training and tools required to provide impactful responses to keep children in school. This has allowed for improved decision-making and support at the school level, at the same time ensuring teachers had the holistic supports they needed to create effective responses. 

At the same time, there was a strong focus on interweaving SiPKPM into government systems from the very beginning. Taking a human-centred, equity-driven approach, the AI models were co-developed across eight workshops involving 30 data experts from seven Ministry divisions. Through the six-month pilot, 670 principals, teachers, counsellors and officials were consulted and trained – not simply handed a tool and told to use it. 

It is now operational across more than 10,000 schools nationwide and embedded in cross-divisional government workflows.  

The system goes beyond identifying students at risk of dropping out. It also asks an important question: once students are re-engaged, how do we keep them in school? 

It includes adaptive learning pathways that personalize learning content and academic guidance based on a student’s performance, personal traits, and socio-economic background. It also offers AI-supported career guidance, with recommendations used by teachers and school counsellors to help students understand the vocational, academic, and alternative routes available to them – matched to their individual strengths and aspirations.

This is a meaningful shift. Rather than simply plugging students back into a system that may have already failed them, SiPKPM seeks to make school relevant and navigable for each individual child. By aggregating patterns in attendance, performance and risk factors, it creates a structured feedback loop that informs planning and decision-making at the school, district, state and ministry levels. This links individual support with broader system improvement, enabling more responsive and continuous refinement of education delivery.

What the system ultimately represents is a proof of concept for responsible, evidence-based AI in education. Not technology imposed from above, but built into the fabric of daily decision-making, led by government, trusted by teachers, and designed – always – around the child.

Key innovations

• Predictive analytics: Machine learning models score each student across seven indicators.
• Adaptive learning pathways: AI personalized content and guidance based on student performance, learning behavior and socio-economic context, supporting engagement and academic progress.
• AI-supported career guidance: Evidence-based recommendations to guide students toward vocational, academic and alternative pathways aligned with their strengths and aspirations.
• System integration: Structured monitor-intervene-monitor cycles embed AI insights into workflows across schools, district/state offices, and the federal Ministry, linking local actions to national policy priorities.
• Capacity building and co-development: Ensuring local relevance, policy alignment and government ownership.
• Iterative refinement: Pilot across six zones nationwide informed dashboards, intervention triggers and workflows to improve responsiveness and contextual relevance.
• Generative AI intervention module (2026): Completion and rollout of a generative AI module to support teachers, counsellors, schools, district and state offices in analyzing risk data and informing intervention decisions.