Using AI to Reduce Fraud in Federal Programs
Published: October 04, 2018
Artificial intelligence (AI) holds great promise for improving government efficiency and effectiveness through its capacity to automate tasks, improve citizen services, increase security, and combat fraud. However, government agencies are only in the early stages of exploring and adopting AI and machine learning technologies.
The House Oversight and Government Reform IT Subcommittee recently conducted a series of congressional hearings on AI to promote awareness and adoption in the federal government. One area exhorted for AI’s potential was in reducing waste, fraud and abuse.
Federal improper payments totaled $141 billion in FY 2017. These payments include those that should not have been made, were made in an incorrect amount, to eligible recipients, or for ineligible goods or services. Reducing improper payments is a high priority for federal agencies and the administration.
Rep. Will Hurd (TX-R) in one of the AI committee hearings cited a 2016 GAO finding that Medicare overpayments totaled $60 billion, and “if AI was used to identify those overpayments faster, investigators could have focused on the costliest overpayments first and saved money.”
Ian Buck, vice president and general manager of accelerated computing at NVIDIA and in witness one of the oversight hearings, gave the example of PayPal using AI to detect credit fraud, which has cut their fraud rates by 50%.
Clearly, the federal government has much to gain from implementing AI to reduce waste, fraud and abuse, but few agencies are in a production mode to date. AI is garnering much interest among agencies and a number of proof-of-concepts and pilot projects are underway in areas of customer service, medical research, health care and procurement innovation. However, I was unable to locate many examples of AI currently in production or pilot testing for purposes of fighting waste, fraud and abuse.
One example I did find was a 5-year contract signed in May between CMS and NCI to reduce improper payments in Medicaid and CHIP claims. NCI will use a range of technologies such as advanced analytics and AI on the $44 million contract to detect inappropriate spending.
Another example is IRS’ quest for cloud-based AI to enhance cybersecurity. IRS released an RFI in July seeking information on a wide range of technologies such as AI, machine learning, cognitive computing, and data analytics techniques and algorithms. These technologies would be applied in cybersecurity areas like threat intelligence, insider threat, cyber operations and processing, exploitation and dissemination, and big data analytics, all of which would protect against theft of data that might result in tax fraud.
Additionally, a recent GCN article stated that the IRS is also exploring machine learning to improve agency business practices related to identity, refunds and fraud.
The market remains ripe for application of AI to reduce waste, fraud and abuse. However, agencies are only in the early stages of taking advantage of this technology. Contractors can look for more opportunities in this area as pilot projects bare positive results and IT modernization efforts progress.