Intelligence Engine
Three-layer hybrid: statistical engine → signal classification → AI/RAG briefing. The AI does not make decisions. It only summarises structured evidence.
L1
Statistical Engine
Pure, deterministic calculations from the four datasets.
- ·Latest NEET count
- ·Latest NEET rate
- ·Unemployed NEET
- ·Economically inactive NEET
- ·Inactive share
- ·Quarter-on-quarter movement
- ·Year-on-year movement
- ·Age/sex movement
- ·Release comparison vs previous publication
- ·Sampling variability / confidence interval
- ·Trend persistence (consecutive QoQ rises)
L2
Signal Classification Engine
Transparent if-then rules. Every signal lists why it triggered, with formula and evidence.
- ·Participation Pressure
- ·Inactivity-Driven Pressure
- ·Labour-Market Absorption Risk
- ·First-Rung Access Risk
- ·Hidden NEET Visibility Gap
- ·Never-Worked Signal
- ·Graduate Absorption Risk
- ·Long-Term Detachment Risk
- ·Work-Readiness / Capability Formation Signal
- ·Local Delivery Data Required
L3
AI / RAG Briefing Engine
Language-only layer. Reads the evidence payload and produces a draft briefing. Never invents causes.
- ·Reads only the structured evidence payload
- ·Cites detected signals, confidence band, and source tables
- ·Lists local data required for diagnosis
- ·Always emits draft status — never final without human review
- ·Swappable in-house model slot — current prototype is deterministic
Boundary statement
This system does not use AI to judge young people. It uses AI to help institutions see where the education-to-work participation system is weakening, so support can be targeted before detachment becomes long-term and harder to reverse.