MIRA — Albania's AI Risk-Assessment System for Labour Inspection Targeting

Informal employment has affected more than half of Albania's workforce in recent years, concentrated in agriculture, trade, construction and hospitality. Albania's State Inspectorate for Labour and Social Services (ISHPSH) previously relied on a static rules-based 'Matrix of Penalties' to decide which businesses to inspect. With technical support from the International Labour Organization (ILO) and funding from the European Union, ISHPSH built MIRA (Matrix of Intelligence and Risk Assessment), which runs nine machine-learning algorithms over historical inspection data — business size, sector, location, seasonal patterns, working hours and compliance history across 100+ variables per inspection — to score which employers are most likely to be non-compliant. MIRA went live in November 2023, replacing the previous rules-based system.

With more than 12,000 inspections conducted annually, 70% are now planned using MIRA's risk assessment (the remaining 30% stay random, to check the model isn't missing blind spots). ISHPSH and the ILO report a 30% improvement in predicting undeclared and underdeclared work compared to the previous manual method, and about 68% of confirmed informal-employment cases now come from MIRA-planned inspections rather than random ones. No independent third-party audit of MIRA's fairness or error rates has been published.