AI Is Keeping Watch Over Government Spending




As the world turns increasingly more digital and data-driven, there is an increasing desire for greater visibility and transparency of data. Governments around the world have turned to digital means to submit and pay taxes as well as collect a variety of revenue from different sources. Likewise, governments are making deeper use of data and systems for their expenditures and analyzing the patterns of that spending. 

One of the lesser-known agencies in the US federal government is the Bureau of the Fiscal Service (BFS). As a bureau of the U.S. Department of the Treasury, the BSF manages the federal government's accounting, central payment systems, and public debt. In essence, the BFS is the bookkeeper for the US federal government. A huge role given the trillions of dollars that flow through US coffers on an annual basis. Since the Federal Funding Accountability and Transparency Act of 2006 (FFATA) was signed into law on September 26, 2006, the BFS has embarked on a number of wide ranging data-centric efforts to provide visibility into government spending including USASpending.govFiscalData.Treasury.gov, and DataLab.USASpending.gov

Not surprisingly, the BFS has also invested heavily in the use of AI, the main topic of an upcoming AI in Government presentation on November 18, 2021  with Justin Marsico, Chief Data Officer of the Bureau of the Fiscal Service. In that presentation, Justin shares how deeply the bureau is investing in the use of AI and some of the ways in which it is providing insights into government spending and revenues.

In this article, Justin shares some of the insights of AI use in the bureau. 

What are some innovative ways you’re leveraging data and AI to benefit the Bureau of the Fiscal Service (BFS)? 

Justin Marsico: At the Bureau of the Fiscal Service, we are using AI to innovate complex financial data for the public. For example, AI helps convert appropriation bills into structured data so we can assign spending data to specific programs. This helps us save time by making the process easier through data visualizations, and other visual tools. We are also beginning to experiment with AI for fraud detection and neural search to improve the search function on our public facing websites. 

Justin Marsico: Modernizing data storage into the cloud provides opportunities to automate ETL (extract, transform and load), data quality assessments, outlier detection and improves our ability to set the foundations for production and machine learning systems. 

How do you identify which problem area(s) to start with for your automation and cognitive technology projects? 

Justin Marsico: We work closely with our federal process improvement team to identify which processes can benefit from robotic process automation compared to processes that would benefit from artificial intelligence. The use cases for AI often come from process improvement efforts and techniques. There are also well-known use cases from innovation in banking, accounting, and data collection we have learned from, and can apply to our current systems. 

What are some of the unique opportunities the public sector has when it comes to data and AI? 

Justin Marsico: Broadly, the public sector has great opportunities in data synthesis, data fusion, and data collection to provide the public with information to build critical systems that expedite measuring policy effectiveness, public safety, and optimize open data sharing. In the Bureau of the Fiscal Service there are great opportunities to help build systems that federal employees use to increase accuracy in financial reporting and lower lead times for providing monetary services to agencies. 

What are some use cases you can share where you successfully have applied AI? 

Justin Marsico: The Bureau has successfully piloted AI for fraud detection for Revenue Collections Management and has also piloted an effort for natural language processing for congressional appropriations bills into structured data that can be used to assign spending to federal accounts.

Can you share some of the challenges when it comes to AI and ML in the public sector? 

Justin Marsico: Governance and data sharing can always be a challenge both in the private sector and public sector. There are rightfully data ethics concerns about using machine learning in ways that don’t discriminate. There is also a trade off with the shareability of data, IT security and privacy that always needs to be considered. The government is actively modernizing systems and often there’s added lead time while waiting for availability, governance rules, cyber security needs and API development before deploying production machine learning systems. 

How do analytics, automation, and AI work together at the Bureau of the Fiscal Service (BFS)? Justin Marsico: Analytics and AI often go through layers of peer review, so AI explain ability or causal modeling is important. General dashboarding can be automated, but when research questions explore causality with the goal to provide explanations for strategic decision makers, but this type of work will be done with scientific rigor here. 

How are you navigating privacy, trust, and security concerns around the use of AI? 

Justin Marsico: We established a data governance board and an analytics subcommittee that brings together stakeholders to help with the secure development of production systems.  We are also adopting the Federal Data Strategy Data Ethics Framework for specific uses at Fiscal Service. When properly implemented, data ethics will help Fiscal data users make fair decisions and promote accountability throughout the data lifecycle.

What are you doing to develop an AI ready workforce? 

Justin Marsico: We have established a bureau-wide Analytics Community of Practice. In concert with that community, we’re developing a community led certification program for common data science tools. We also just finished a bureau-wide data challenge and plan on conducting that annually, along with exploring the use of other periodic challenges for both the public and our certification program. 

What AI technologies are you most looking forward to in the coming years? 

Justin Marsico: One area we are looking forward to are developments in synthetic data that could help us improve the availability of public data. Also, trends in deep reinforcement learning could have dramatic implications for agency operations. To support AI, we’re always paying attention to trends in data engineering systems including the data mesh architectures and the use of machine learning to help standardize data attributes across disparate systems. Quantum AI systems are on the distant horizon but could also have major implications so as that technology matures our office will continue to monitor its developments. 

Justin plans to dive more deeply into the above topics as well as address further questions on AI and data at the AI In Government online event.

No comments

intech company. Powered by Blogger.