Avoiding the costs of fragmentation: digital transformation in social protection

Blog Digital transformation

Written by Anita Mittal (GIZ), Melis Guven (World Bank), Veronika Wodsak (ILO).

When social protection schemes and programmes speak different languages, the people who need to access benefits and services most pay the price. Digital transformation offers a way forward, but only if inclusion is built in from the start.

When a crisis hits, the speed and reach of a social protection system matter enormously. Yet across many countries, what greets people in moments of urgent need is a patchwork — parallel registries that don’t talk to each other, payment channels that exclude those without formal identification, and different programmes requiring people to navigate separate eligibility rules, application processes, and agencies.

The fragmentation is rarely deliberate. It accumulates over decades of siloed investments, each well-intentioned on its own terms, and each producing data and processes that were never designed to connect or to make programmes easily accessible. The question this piece tries to sit with is not simply whether digital systems can fix this. That framing is too optimistic, and probably too narrow. A more honest question is: under what conditions can digital transformation reduce the human cost of fragmented social protection, and what does it take to get those conditions right?

Fragmentation is not just an administrative problem

It is tempting to treat fragmentation as a technical inconvenience — a matter of incompatible databases and bureaucratic silos that better software could resolve. The reality is considerably more stubborn.

Fragmentation produces exclusion. When a person’s name is spelled differently across three registries, or when a programme’s eligibility criteria require documentation that an informal worker simply cannot provide, the system’s inefficiency becomes that person’s crisis. Research on social protection coverage has consistently shown that the “last mile” problem — reaching the most vulnerable populations — is not only a resource challenge; it is also fundamentally a delivery systems and design challenge (Barca and Chirchir, 2014). These design choices ultimately shape people’s ability to access and benefit from social protection.

Fragmented architectures tend to reinforce existing inequalities: those with formal employment histories, stable addresses, bank accounts and government-issued documents navigate multiple programmes with relative ease; those without these markers are filtered out repeatedly, at every point of interface.

There is also a fiscal dimension that often goes unexamined. Parallel systems require parallel investments — separate staff, separate infrastructure, and separate monitoring mechanisms. Countries with limited fiscal space can be paying a “fragmentation tax”, spending more to deliver less to fewer people. The cost of not investing in greater integration and interoperability is real, and it compounds over time.

What digital transformation can realistically offer

The promise of digital systems in social protection is significant, but it deserves careful treatment. Interoperable social protection information systems, digital payment infrastructure, and unified case management systems have, in several documented cases, expanded coverage, reduced processing times, and lowered delivery costs substantially (Leite et al., 2017). These are not trivial gains. When a single application can trigger eligibility checks and referrals across multiple programmes, or when a beneficiary update propagates automatically rather than requiring repeated visits to separate offices, the practical difference in people’s lives can be considerable.

At the same time, the evidence base is more mixed than the technology discourse sometimes acknowledges. Successfully increasing coherence across schemes and programmes requires not only technical interoperability solutions but also has a political dimension: different entities have to collaborate and agree on exchanging data and information, as well as joint processes, responsibilities and accountability structures that facilitate this coordination. Also, digital systems inherit the biases embedded in the data on which they are built. If the underlying data and registration processes have historically excluded informal workers, migrants, women, people without identification, or other hard-to-reach groups, digitizing the system does not automatically extend its reach to them. In fact, automation can sometimes accelerate exclusion, processing ineligibility determinations faster than any appeals mechanism can absorb them.

This tension sits at the centre of the Digital Convergence Initiative (DCI) — the joint effort by the Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ), the International Labour Organization (ILO), the World Bank, Expertise France (EF), and FIAP to support countries in building more coherent and interoperable digital social protection systems. Funded by the German Federal Ministry for Economic Cooperation and Development (BMZ) and the European Union, the DCI brings together complementary expertise across its five partners to advance shared standards, institutional capacities at national levels, country-level reforms, and a global knowledge agenda on digital social protection.

The DCI framework does not treat digital transformation as a destination. It treats it as a lever for extending social protection coverage — one that needs to be positioned carefully relative to a country’s institutional context, data governance environment, and the populations it is trying to reach. Through its AI Hub, the DCI is also beginning to examine how artificial intelligence tools can build on stronger digital foundations to improve service delivery and responsiveness — while ensuring that responsible governance and appropriate safeguard frameworks accompany their use.

Inclusion as a design imperative, not an add-on

One of the more persistent mistakes in digital social protection projects is treating inclusion as something to be addressed after the core system is built — a supplementary module, a communications campaign, or a grievance mechanism bolted on at the end. By that point, the fundamental architecture has already made its choices.

Inclusive digital systems require inclusion to be specified as a constraint from the outset. This means designing for the person who does not have a smartphone, who does not read, who lives three hours from the nearest agent banking point, and whose identity may exist only in community records rather than civil registration systems. It means asking, repeatedly and at every design stage, whom this system will not serve and why.

Some countries have made meaningful progress here. The use of alternative identification pathways, community-based verification mechanisms, and assisted digital enrolment has demonstrably extended coverage in contexts where strict digital ID requirements would have created large-scale exclusion (World Bank, 2021). These approaches require more institutional effort and coordination than a purely digital-first design, but the coverage gains tend to be substantially larger.

Experience from DCI’s country engagements also points to the importance of engaging intended users, civil society organisations, and community representatives in the design and implementation of digital systems, particularly where systems need to reach populations at risk of exclusion. When those closest to excluded populations have a voice in shaping eligibility criteria, enrolment processes, and grievance channels, the resulting systems are both more inclusive and more trusted.

Responsiveness: the test that most systems still fail

A social protection system’s responsiveness — its ability to scale up rapidly during a shock, and to reach people who were not previously registered — is perhaps the starkest measure of whether digital investment has been well targeted.

The COVID-19 pandemic offered an involuntary stress test on this dimension, and the results were instructive. Countries with pre-existing interoperable information systems for various programmes and established digital payment infrastructure were generally able to extend emergency transfers more quickly and to broader populations than those without (Gentilini et al., 2020). Yet even among the faster responders, significant gaps emerged. Urban informal workers, migrants with irregular status, and people experiencing homelessness were frequently missed by systems whose adaptive mechanisms had not been designed with those populations in mind.

Responsiveness, it turns out, is not simply a function of system speed. It depends on the quality of data already in the system, the institutional arrangements for cross-agency coordination, and the legal frameworks governing data sharing in emergency contexts. Digital infrastructure is necessary but not sufficient. The pandemic made visible what practitioners in social protection had long argued: that a system is only as responsive as its weakest institutional link.

The deeper case for integration

What the fragmentation debate ultimately reveals is a question about what social protection systems are for. If the goal is to deliver benefits efficiently to populations that are already enrolled and whose circumstances are broadly stable, fragmented systems can, with enough investment, be made to perform adequately. If the goal is to provide genuine protection across a variety of life cycle or covariate risks — to be present for people at the moments when their lives are most precarious — then fragmentation imposes costs that no amount of efficiency within individual programmes can offset.

Integrated digital delivery systems, when designed with inclusion as a non-negotiable constraint, offer a different kind of possibility. They allow countries to move closer to the idea of a social protection floor for all members of society— a set of guaranteed minimum protections that are accessible to people regardless of which programme they happen to qualify for, which agency they first encounter, or whether their circumstances fit neatly into a predetermined category. Emerging tools, including AI-powered analytics, offer additional means to identify coverage gaps and strengthen referrals — but only when grounded in sound data governance and human oversight. This is not a vision that digital technology alone can deliver. It requires political will, sustained institutional investment, data governance frameworks that protect individuals while enabling responsible information sharing, and the kind of long-term cross-agency coordination that is genuinely difficult to sustain.

Some countries have started to transform their social protection systems with a human-centric perspective as their starting point: ensuring that data is shared across the publication so that people are only asked for it once, and introducing automatic triggers for certain benefits or support services upon a certain life event such as the birth of a child or the onset of a disability. The DCI has seen this clearly: the technical architecture is only one part of the challenge. Institutional arrangements, political commitment, and sustained cross-agency coordination are critical for success.

The cost of fragmentation is ultimately borne by people, not systems. Acknowledging that cost honestly, and investing in the structural changes needed to reduce it, is what it means to take social protection seriously and build systems that work for everyone.

References

This blog is a contribution from the Digital Convergence Initiative (DCI), a joint programme of GIZ, ILO, the World Bank, Expertise France (EF), and FIAP, focused on building interoperable, inclusive, and adaptive digital social protection systems. The DCI is funded by the German Federal Ministry for Economic Cooperation and Development (BMZ) and the European Union.

This blog post was originally published on socialprotection.org as part of the The Cost of Not Investing in Social Protection blog series that commemorates socialprotection.org’s 11th anniversary campaign. The series brings together contributions from the USP2030 Working Groups and other invited authors, combining broader reflections on financing social protection with discussions examining the costs of underinvestment across key cross-cutting areas, including gender and inclusion, climate and crises, digital transformation, and food security. Together, the contributions highlight the human, social, and economic consequences of insufficient investment and reinforce the campaign’s central message: the greater cost lies in failing to invest in social protection.