This article explores how digital technologies, including AI, are enhancing the delivery and inclusiveness of social protection systems – while also highlighting the risks, governance challenges, and equity considerations they raise. Drawing on global examples from Brazil, Moldova, Korea and others, it offers lessons for harnessing innovation without leaving anyone behind.

Digital transformation and real time data sharing through interoperable information systems contribute to enhancing the inclusiveness and accessibility of social protection by streamlining enrolment, eligibility verification and service delivery. Artificial intelligence (AI) offers opportunities to achieve these goals but also poses risks that need to be tackled. 

Digital technologies help reduce coverage gaps and reach the right people, at the right time with the right support. The high impact of digital technologies on several measures of universal social protection performance was evidenced during the COVID-19 pandemic. For example, in Brazil, the Emergency Grant app received 57.2 million requests in 2020, a record number of annual enrolments (Alfers and Juegens-Grant 2023)[1].  An analysis of 53 cash transfer programmes showed that countries using electronic systems delivered on average their first payment a month before those relying on manual mechanisms during the pandemic (Beazley et al. 2021). 

Digital services innovation allows to cover workers in specific circumstances. For instance, ILO supported the Government of Moldova to digitize the payment of social security contributions from seasonal agriculture workers, who are mostly women, thus encouraging their formalization and contributing to increased gender equality in social protection coverage.

Digital transformation of social protection also improves efficiency and reduces administrative costs, allowing resources to be redirected toward expanding benefits and reaching underserved populations. Robotic process automation for repetitive tasks can reinforce the capacity of social protection institutions to respond to increased workloads. For example, by automating 14 processes, the Korea Worker Compensation and Welfare Service (COMWEL) saved an average of 26 minutes per claim, resulting in an annual total workload reduction equivalent to 41 full-time employees (ILO 2024). 

Today, artificial intelligence (AI) is increasingly being used in the service delivery and administration of social protection. Natural language processing (NLP) enables social security institutions to extract valuable insights from beneficiary inquiries, complaints and historical records, which help brings service delivery and process automation to new levels. 

For instance, Estonia’s employment service profiles claimants of unemployment benefits and matches their profile to local job offers, by way of recommendations to case-workers. AI agents learn to improve automation processes, which capture and categorize information from grievances stored in hotlines and help prioritize remedial actions. If well designed and implemented, such solutions can enable cost savings and potentially increase precision and consistency of decision-making in a context of heavy data workloads, thus increasing the institutional capacity to extend social protection.

While the benefits of digitalization of social protection should not be overlooked, it is important to consider the limits, risks, costs and potential of digitalization for causing additional disadvantages, especially with the use of new foundational technologies such as AI.

First, some poor countries and large segments of their populations still face barriers to digitalization due to low availability, accessibility and affordability of digital technologies. Low internet reach and cellular footprint persist in many regions. 

Secondly, many social security organizations must first invest in getting control of their data. Data quality and its protection are a critical foundation to the success of digital transformation. Data driven decision-making (rules or AI assisted operational decisions, analytics, and forecasting), and robotic automation processes, rely critically on the quality of the underlying structured and unstructured data.  To that end, ILO has been supporting the National Social Security Fund in Kenya in data quality improvement, through data stewardship and effective data management processes. 

As digital data is increasingly produced, exchanged and consumed, data from social protection organizations also needs to be trusted from a security standpoint. Social protection institutions, notably health insurance bodies, are increasingly threatened by misuse of personal data by third-party vendors and data theft including ransomware. Unfortunately, modern data protection regulations and principles are not always present or enforceable in countries that are undergoing often advanced digitalization of social protection systems.  

Thirdly, it is essential to affirm the primacy of human rights and principles in digital transformation and technology adoption. The ILO contributes, with other multilateral and bilateral organizations, to affirm the relevance of international social security standards and to make sure that universal access, adequacy, inclusion, gender equality, non-discrimination, transparency, fairness, participation, quality of administration and services, accountability, effectiveness, universality and sustainability are at the forefront of modernization and technological improvements (ILO 2023, ILO, ISSA and OECD 2025).  

Social protection institutions leveraging AI need to align their technological investments with national social protection policies and strategies and human rights principles. A non-digital option, including human guidance to access digital interfaces, should always be available so that those who cannot access/use the digital options are not excluded. Digital interfaces should use accessibility technologies (simplified language, use of voice for blind, etc.).  

AI-generated content, either through natural language assistants or support materials that guide beneficiaries to access social protection, need to be adapted to local languages, customs and cultural values to ensure accessibility. 

Social protection agencies also need to increase the representation of groups who are digitally disadvantaged who tend to be under-represented in data sets and statistics that train AI models and consider implementing techniques to mitigate bias to ensure fair outcomes for all beneficiaries. 

Finally, the adoption of digital technologies should contribute to increased levels of trust and confidence in social protection. This includes the appropriate use of AI in different stages of the social protection delivery cycle, guided by risk assessments, the need for human oversight and accountability to review, validate and adjust partially or fully automated decision-making systems, and enhanced grievance and redress mechanisms (Lee-Archer 2023, Ohlenburg 2020). 

ILO social security standards emphasize the need for the participation of representatives of workers and employers, as well as other stakeholders, in the formulation of social protection policies and strategies, including about the use of digital technologies for social protection. Ultimately, social protection institutions should have the capacity to review and reject digital solutions that do not meet the needs of beneficiaries, and beneficiaries should have the opportunity to seek recourse if they can’t access these solutions.

To raise awareness about the potential of digital transformation, data exchange and new foundational digital technologies, including AI, the ILO is partnering with other international actors, in the scope of the Social Protection Inter-Agency Cooperation Board (SPIAC) group on Digitalisation, and the Digital Convergence Initiative Digital Convergence Initiative - DCI - Social Protection, which will also help to produce evidence for a more impactful, respectful and reasoned use of digital technologies in social protection.

[1] In Namibia, also through SMS, the government received almost 80 per cent of applications in one week, while in Peru more than 3 million households were registered during the two-week online registration window (Barca & Hebbar 2020). 

References

• Alfers, L. and F. Juergens-Grant. 2023. “Social protection, the COVID-19 crisis, and the informal economy Lessons from relief for comprehensive social protection”, WIDER Working Paper 2023/93, UN-WIDER and WIEGO. s
• Barca, V. and M. Hebbar. 2020. On-demand and up to date? Dynamic inclusion and data updating for social assistance. Berlin: Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ).
• Beazley, R., M. Marzi, and R. Steller. 2021. “Drivers of Timely and Large-Scale Cash Responses to COVID19: What Does the Data Say?”, Social Protection Approaches to COVID-19 Expert Advice Service (SPACE). London: DAI Global UK Ltd.
• ILO. 2023. “Risks and Benefits of Digital Tools for Social Protection Delivery from a Gender Perspective”, Expert paper prepared by: Becky Faith, Institute of Development Studies, University of Sussex, UK.
• ILO. 2024. “Recent advances in the digitalization of employment injury insurance administration Evidence from Indonesia, Malaysia and the Republic of Korea”.
• ILO, ISSA and OECD. 2025. Social security and digitalization for an inclusive future of work, Paper prepared under South Africa’s G20 Presidency.
• Lee-Archer, B. 2023. “Effects of digitalization on the human centricity of social security administration and services”, ILO Working Paper No. 87.
• Ohlenburg, T. 2020. AI in Social Protection – Exploring Opportunities and Mitigating Risks. Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) and Asian Development Bank (ADB). Bonn.