Skip to main content
Industry

How Microsoft and AI work to reduce government benefit fraud and error

An organized criminal group defrauds a national stimulus program through the use of elaborate fake identities. An 87-year-old woman receives maternity benefits, decades after giving birth. A newly unemployed worker unwittingly submits incorrect online forms and is approved for twice the benefits he qualifies for. These representative anecdotes are just a sampling of the myriad ways in which funding earmarked for government benefits can be lost to fraud, abuse, and waste.  

Governments do remarkable work to help people live with dignity and endure difficult circumstances by providing essential benefits and resources across a broad spectrum of programs. Yet the challenge of providing the right benefits at the right time to all qualified recipients while also minimizing the incidence of improper payments is a balancing act that many governments struggle to achieve. 

Helping agencies and organizations to address these types of challenges is central to our work in Microsoft for Government. We focus on enabling thriving communities and inclusive programs through technology. In the case of benefits protection, we work to help each government discover the right mix of technological and organizational innovation to mitigate risk within the parameters of their unique circumstances and requirements. 

Microsoft for Government

Help solve society’s biggest challenges.

A red-haired woman with a green shirt and glasses, sitting inside an office working on her laptop. She has a cup of coffee or tea and a phone by her side.

The rising tide of improper payments

The loss of benefits funds to unauthorized payments has long been a problem for governments around the world. In recent years, the scale and complexity of the challenge has escalated. Fraud is a particularly expensive and multidimensional problem. To offer just one example, the US Government Accountability Office (GAO) estimated that the amount of fraud in US unemployment insurance programs during the COVID-19 pandemic was likely between $100 billion and $135 billion USD.1

The reasons for this are varied, and they parallel larger trends in society. A key driver is the ready availability of AI tools, which fraudsters have enthusiastically embraced to automate, scale, and generally uplevel the impact of their efforts. But it’s not always a matter of criminal intent. So-called “unintentional errors”—mistakes made without maliciousness due to things like administrative errors or poorly designed solutions—also incur huge costs. According to the GAO, overpayments or payments that should not have been made (for example, to deceased people or those no longer eligible for government programs) totaled an estimated $247 billion during fiscal year 2022.2

Why governments struggle to ensure proper payments 

Unlike organizations in other global industry sectors, governments face unique challenges in administering complicated social benefits programs. The nature of distributing funds accurately and efficiently to constituents in all corners of society is inherently fraught. Governments must contend with a set of vexing factors, including: 

  • Growing demand for benefits funding—A mix of aging populations, increases in immigrant and refugee populations, and people seeking broader ranges of benefits is driving higher budgets and greater complexity. 
  • Mandates to “do more with less”—Austerity programs and budget cuts are colliding with rising public expectations for modern services and impacting the ability of security teams to procure top cybersecurity talent.  
  • Disconnected data silos—Across departments and agencies, teams often work with older tools and isolated systems that cannot share data easily or securely, making integrated new “tell us once” solutions difficult to implement. 
  • Ineffective legacy systems—A high “technical debt” of outdated computing environments and applications impedes innovation and threatens to hinder the most vulnerable people from accessing payments they’re entitled to.  

Beyond these costs and considerations, the larger risk here is the erosion of public trust. Especially in social benefits systems, which exist to ensure the welfare of people across all walks of life, government is obligated to demonstrate reliability and integrity. Widespread fraud and abuse do more than increase deficits and siphon off critical resources. They also undermine confidence in all aspects of government.  

A vision for modern benefits delivery and protection 

Imagine if governments were able to innovate on par with private industry. Imagine if people could easily access all the information they need through any channel (online, over phone, text, and more) and have their own virtual assistant to help them apply for benefits using natural, everyday language. Imagine if these systems would inherently enforce rules and keep up with constituents’ changing lives. Imagine if customer service representatives had access to a complete, 360-degree view of the constituent—if it knew their history and all their previous contacts with every social service agency. And imagine that such things could be accomplished without breaking the budget.  

These are obviously lofty, aspirational visions. However, thanks to the relatively recent technological advancements, some innovative government social service organizations are doing just that. The advent of generative AI and the innovation it has inspired has already delivered promising results on many fronts of government operations, including benefits. To cite just one example, a generative AI-powered chatbot developed in India called Jugalbandi is helping people get assistance for any of 171 government programs, simply by conversing through mobile devices in 10 of India’s 22 official languages. “This is revolutionary for people who could not interact with tech because of language barriers,” said Abhigyan Raman, a project officer. 

Ensuring the integrity of benefits payments means reevaluating a government’s existing slate of technology investments and charting a multistep course of transformation. Embracing AI is a common theme, but the destination in terms of functional outcomes and benefits will be unique to the organization.  

Three areas of generative AI impact in benefit payments protection

In our work with customers who are either early in the journey or simply just embarking, we identify three categories of solutions that hold the greatest promise for cloud and AI innovation:

  1. Boost workforce productivity. Governments can potentially save time and money using generative AI for scenarios such as live transcriptions and translations during benefit eligibility interviews. AI can also draw insights more easily from large volumes of data. One department, for example, identified GBP14 million in potentially fraudulent loans by analyzing a set of 250 networks of people, organizations, and places, processing 100 million data items.
  1. Embrace “Prevention by Design. AI can integrate controls into systems to detect fraudulent or erroneous activities and enable real-time profiling and alerts. For example, a leading bank is using Voice ID analysis for incoming phone calls from 31 countries and 15 languages, checking more than 100 behavioral and physical vocal traits in a matter of seconds. Since launch, it has flagged more than 43,000 fraudulent calls and reduced fraud by 50%, preventing an estimated GBP981 million in losses.
  1. Enhance citizen engagement. When chatbots become knowledgeable assistants, when a person’s voice becomes their password, when the disabled have equal access to resources, then everyone will benefit. For example, a government department is developing an Intelligent Website AI Assistant to help taxpayers more easily process their tax returns. It uses natural, everyday language to address queries and disseminate timely and consistent information. This will help improve compliance and avoid revenue leakages from fraudulent activities.

Beyond technology: How Microsoft partners for the long term 

Embracing the paradigm shift of generative AI obviously begins with technology. The table stakes of modernization for government include migrating to a modern cloud platform and the adoption of a comprehensive AI development solution from a vendor who demonstrates a deep commitment to security and responsible AI practices

Microsoft invests heavily in all these areas. However, success involves much more than just technology. Governments also depend heavily on the contributions of trusted solution providers, and we believe our global partner ecosystem sets us apart, with expertise in all corners of the world. Then, the final unique benefit we offer is the deep experience of our industry advisors and the many highly experienced government veterans on the Microsoft Government team. 

Our job is to help build the bridge between the technical and the strategic, on realistic terms. When we sit down with customers, we help clarify challenges and goals, educate on important challenges (for example, how governments can tackle cybersecurity and AI skilling), and share our experiences with other governments facing similar challenges (sometimes even connecting them to help foster learning). Then, we embark on identifying and exploring use cases, evaluating impact, and taking the knowledge gained for further innovation.  

We are excited to work with governments to mitigate fraud and abuse in payments. To learn more about how Microsoft is helping to create opportunities that support vulnerable communities, see our Microsoft public health and social services website, and learn more about Microsoft for Government


1Estimated Amount of Fraud during Pandemic Likely Between $100 Billion and $135 Billion, GAO. September 2023.

2Federal Payment Errors, Known As Improper Payments, Are A Continuing Concern, GAO. March 29, 2023.