From Benefits to Better Jobs: How AI Could Reshape Employment Support in Belgium and Greece

AI-powered employment tools could help Belgium and Greece move minimum-income beneficiaries into more stable jobs by improving skills matching, career guidance and access to training. Success will depend on stronger data integration, human oversight, employer participation and wider reforms that create sustainable, higher-quality employment.

Artificial intelligence could become an important bridge between social protection and employment in Belgium and Greece, but technology alone will not solve long-term unemployment. An OECD study, developed in cooperation with the European Commission's Reform and Investment Task Force and financed by the European Union through the Technical Support Instrument, finds that both countries already have strong digital foundations. The bigger challenge is connecting welfare, employment, skills, and labour-market information so Minimum Income Scheme (MIS) beneficiaries can move into stable, better-paying jobs.

The issue matters beyond social protection. Helping more beneficiaries enter sustainable employment could expand labour-force participation, reduce long-term pressure on public finances and give employers access to workers who may currently be overlooked. For policymakers and development partners, the report shows where investment in AI can help, and where broader reforms remain essential.

Employment Numbers Reveal a Persistent Gap

Around 1.6% of Belgium's population depended on the MIS at the beginning of 2024, after the number of beneficiaries more than doubled between 2003 and 2023. Greece had approximately 240,000 Guaranteed Minimum Income beneficiaries across 172,000 households at the end of 2024.

The employment gap is substantial. In Belgium, more than half of new MIS entrants remain on the scheme one year after registration. Only around 13% move into regular employment within a year, compared with nearly 40% of unemployment-benefit recipients.

Greece shows a similar pattern. Around seven in ten jobseekers registered with the Greek Public Employment Service, DYPA, find employment within two years, compared with only three in ten MIS beneficiaries.

Skills are part of the problem. About 49.1% of Greek beneficiaries have less than upper-secondary education, compared with 19.7% of the general population aged 25-64. Around 73.9% reported being unemployed when first applying for the benefit.

These numbers suggest governments need to move beyond simply registering beneficiaries with employment services. Support needs to address qualifications, skills, individual employment barriers and differences in local labour demand.

Finding Work Is Only Half the Challenge

The report highlights another important policy problem: getting someone into a job does not necessarily mean they have achieved sustainable employment.

In Greece, only 38% of beneficiaries entering employment remain continuously employed for the following year. More than half experience at least one occupational change during the three years following their first post-GMI job. Many find work in hospitality, cleaning and personal services and are more likely to have part-time or shift-based jobs.

Belgium's Article 60 programme also illustrates the challenge. It provides temporary subsidised employment to disadvantaged beneficiaries, but three years after entering unemployment benefits, only around one-third of former Article 60 participants are employed, compared with roughly half of those without previous Article 60 participation.

Around 70% of employment outcomes among former Article 60 participants are concentrated in manual occupations, compared with approximately 45% among non-participants.

For policymakers, this means success should be measured through job retention, earnings and career progression, not simply the number of people placed in jobs.

AI Can Connect Skills, Vacancies and Career Pathways

The OECD proposes an AI-supported platform centred on five functions: personalised job matching, better vacancy management, identification of skills and employment barriers, labour-market intelligence, and recommendations for career progression.

Instead of simply showing available vacancies, such tools could identify transferable skills, suggest alternative occupations and recommend training when beneficiaries lack required competencies.

Both countries have a head start. Belgium already uses sophisticated digital and AI tools for profiling, skills identification, vacancy matching and career guidance. Greece's DYPA provides more than 80 e-services alongside profiling, digital counselling and emerging AI applications.

The main weakness is fragmentation. Information is spread across welfare agencies, employment services, municipalities and administrative databases.

Belgium could connect its regional employment systems through a federally coordinated interoperability framework. Greece could strengthen links between the MIS registry, DYPA, ERGANI employment records and active labour-market programme data.

For development partners, the lesson is that financing should not focus only on purchasing technology. Investment is also needed in data systems, interoperability, institutional capacity, staff training, digital inclusion and rigorous evaluation.

Private Sector Opportunity Comes With New Risks

Better matching could help employers identify candidates whose skills are overlooked by conventional recruitment. This could be particularly valuable where businesses face labour shortages while disadvantaged groups remain unemployed.

Employers also have an important role in supplying accurate vacancy information. Without sufficient participation from businesses, even sophisticated AI systems will have limited opportunities to recommend.

There are risks. Poor-quality or historically biased data could reproduce inequalities. People with limited digital skills could be left behind, while excessive dependence on automated recommendations could weaken professional judgement.

The OECD therefore recommends gradual implementation, human oversight, transparent governance, stronger data protection and compliance with national and European requirements, including the EU AI Act. Governments should continuously measure uptake, recommendation quality, employment outcomes and possible differences across demographic groups.

The broader message for Belgium, Greece and development partners is straightforward: AI should support labour-market reform, not substitute for it. Technology cannot create quality jobs or eliminate every social and skills barrier. But when combined with training, social services, capable counsellors and better labour-market data, it could help governments use resources more efficiently, businesses reach a wider talent pool and vulnerable jobseekers build sustainable careers rather than moving repeatedly between benefits and short-term employment.