Even in the world’s most advanced social protection systems (systems that include contributory social insurance and non-contributory social welfare), there are gaps in the quality, efficiency, and responsiveness of social programs. The Organisation for Economic Co-operation and Development (OECD) Risks that Matter Survey shows that close to half (46%) of people across 27 OECD countries think that they could not easily access social benefits if they needed them. Of those who doubt they could access benefits, over three-quarters (77%) expressed concerns that the application process would be difficult and time-consuming, markedly outweighing concerns about eligibility (57%) or fairness (53%).
Improving the ease and speed of accessing benefits is key to government efforts to extend social and economic safety nets to what the International Social Security Association (ISSA) refers to as the “missing middle”. Self-employed and gig workers, as well as rural, migrant, and domestic workers are typically time-poor, not already engaged in social protection systems, and are often not included in targeted outreach programs. This makes them vulnerable to economic shocks and cost-of-living increases that can tip them into poverty and homelessness.
As such, many government agencies and not-for-profit organisations are looking at ways to make social services more accessible and responsive by reducing the “hassle costs” associated with claiming benefits.
How AI can help
Governments around the world have been realising significant efficiency gains through applying artificial intelligence (AI) in the back-office to improve workforce productivity. Encouragingly, there are also recent examples of AI being leveraged in the front-office to improve the efficiency and effectiveness of citizen engagement.
- Quicker time to payment with AI-supported assessment processes
At Hamburg’s Ministry of Finance, a combination of Machine Learning (ML) and Generative AI (GenAI) support staff to efficiently process applications for more than €3.5 billion in financial aid.
SAP Machine Learning is used to link citizen application data to supporting documentary evidence, enabling case workers to expedite processing for the bulk of applications and to focus their attention on those most likely to be non-compliant. Across two programs, Hamburg reports that nearly 180,000 benefit applications have been processed, with more than 10 million pages of supporting documents automatically evaluated and classified by AI.
SAP Generative AI Hub has also been introduced to summarise inbound applications and to generate draft outbound correspondence, further reducing time to payment for customers while minimising the burden of repetitive manual work for staff.
- Improved customer service with AI-powered workflow automation
Similarly, Spain’s Public Employment Service uses SAP AI to automate workflows and to recommend potential benefits based on customer circumstance data. AI has contributed to a 20% increase in user productivity, which amounts to a substantial efficiency gain when applied to an agency of 9,000 public employees managing €1.8 billion in monthly payments. These efficiencies flow through to citizens, as described by the Deputy Director General of Benefits and Subsidies, who reports their AI-enabled system “…has empowered me to shift focus from administrative tasks to truly enhancing citizen service, allowing for quicker responses and more meaningful interactions.”
- Faster query resolution with AI-enabled chatbots
At Germany’s Federal Foreign Office, an SAP AI chatbot responds to 50% of citizen inquiries with no human intervention, resulting in 77% being answered and closed within the same day. While social services inquiries would typically be more complex, there’s certainly potential for AI to categorise and prioritise inbound communications and to route them to the appropriate channel or group. This is the case for more than 83% of the inquiries being received by the Office every day, which embassy staff say “…means efficient communication, satisfied customers and a gain in personnel resources for other tasks.”
Reducing the barriers to adoption with embedded AI
To date, the types of use cases described above have been delivered as custom AI solutions, limiting uptake to agencies that are sufficiently resourced to assemble AI systems from Large Language Models (LLMs) and other necessary components. This is further exacerbated by the additional work needed to protect customer data, prevent bias, and ensure the reliability of AI recommendations.
Thankfully these barriers to adoption are being reduced as AI capabilities become embedded into enterprise software, enabling agencies to adopt out-of-the-box solutions. For example, something as simple as supporting staff to retrieve information using AI-enabled natural language search could reduce the time customers spend waiting on hold while their case worker struggles to locate their file.
Revolutionising social services with agentic AI
The advent of agentic AI could be a tipping point for social services AI use. By virtue of their ability to take multiple paths and iterative steps towards achieving an outcome, AI agents are particularly suited to the type of complex case processing inherent in social services. We can imagine a future where benefit applications are picked up and processed by a team of specialised AI agents that can autonomously validate compliance, determine eligibility and entitlement, identify potentially fraudulent claims, and present reasoned recommendations to human case workers for approval.
Such a future could be just around the corner. Gartner predicts that, “by 2028, 33% of enterprise software applications will include agentic AI, enabling 15% of day-to-day work decisions to be made autonomously.” Similarly, it notes, “by 2029, agentic AI will autonomously resolve 80% of common customer service issues without human intervention, leading to a 30% reduction in operational costs”.
In summary, AI is already enabling early adopters like Hamburg’s Ministry of Finance to improve the efficiency of application processing for social benefits. AI adoption in social services is now set to scale as AI capabilities are embedded into enterprise software, and this could lay the groundwork for a big leap forward with agentic AI. AI will be increasingly capable of reducing the “hassle costs” associated with claiming benefits, helping to ensure that social and economic supports reach the people that need it, when they need it.
