Software

States Augment Fraud Detection with AI Technology

As bad actors become bolder, agencies must get smarter and they are using artificial intelligence to to do so.

Brian T. Horowitz is a writer and editor covering enterprise IT, innovation and the intersection of technology and government.

Colorado has developed what it calls “the 18th measure,” a way to detect unemployment fraud. Cher Haavind, deputy executive director and chief communications officer for the Colorado Department of Labor and Employment, recently told Bloomberg that the method was instrumental in the identification of more total cases of fraud than the other 17 detection measures deployed by the state.

“CDLE is currently evaluating its full fraud detection and prevention landscape, working with other states and talking to vendors in the fraud field to identify gaps and opportunities,” Haavind says.

Although states now use AI for fraud detection as well as financial management, they are behind the private sector in adopting the technology, according to Darrell West, vice president of governance studies and a senior fellow at the Brookings Institution’s Center for Technology Innovation.

“They don’t have a big IT infrastructure and they don’t have a lot of technical expertise, so it’s been harder for them to figure out how to deploy AI for managing finances,” he says.

“AI is very good at spotting unusual activities, comparing a single transaction to the body of financial transactions that have taken place in a particular agency or state,” West says.

Click the banner below to get access to a customized content experience by becoming an Insider.

Machine Learning Helps Agencies Sniff Out Fraud

Richard Hoehne, lead for IBM’s Worldwide Financial Crimes, says the company works with states to address tax fraud using machine learning (ML).

“We can help identify which returns are the ones you should go audit based on peer-group analysis,” Hoehne says. “When you add machine learning, the process is more precise.”

Supervised ML, like in IBM’s Watson Studio, makes fraud detection possible by learning from past experiences.

“Supervised machine learning is based on the premise that you have known outcomes,” Hoehne says. “You can use these modeling techniques to create statistical representations to find which one identifies known outcomes.”

AI Delivers Significant Value to Agencies in Detecting Fraud

Other tools, such as AutoAI, Data Refinery and low-code visual modeling, help data scientists automate tasks and business users simplify modeling, says Ritu Jyoti, IDC’s group vice president for worldwide AI and automation research.

With a rise in grant fraud, in which grant recipients make false statements on their applications, forge documents and misuse money, users should see acceleration of AI-powered fraud detection solutions, Jyoti predicts.

“By some accounts, when executed properly, AI fraud detection systems can reduce fraud significantly while also reducing the costs associated with detecting them,” Jyoti says. “Results like these provide significant value for government agencies.”

As they start using AI technology for fraud detection, states should start with smaller-scale pilot projects to make sure their algorithms work, West advises. Later, they can scale up from that point.

“Agencies are trying to figure out how to track those transactions and make sure everything is done according to the letter of the law,” West says. “So, I think there will be a large expansion of the use of algorithms for fraud detection at the state level.”

mycola/Getty Images