The more I talk to my colleagues, clients and counterparts about artificial intelligence (AI) and machine learning, the clearer it becomes that the prevention or detection of fraud, waste and abuse is one of the most promising and needed use cases in federal and local government.
The risk is real. Most recently, the Pandemic Response Accountability Committee (PRAC) was established to ensure funds distributed through the Coronavirus Aid, Relief, and Economic Security (CARES) Act are handled properly. However, Inspector General offices have always monitored spending within their agencies and Congress has frequently investigated agency expenditures. Preventing fraud, waste and abuse, or detecting it when it happens, is top of mind throughout the federal government.
I recently participated in a panel discussion for the Government Matters TV program (online replay here), and will take part in a similar discussion for ACT-IAC on July 16. During the Government Matters panel, government and industry experts spoke on the power of the technology, and the use cases the government is adopting today.
Register here for "Machine Learning and Automation: Staying on Track to Detect Fraud, Waste and Abuse," an ACT-IAC panel discussion to be held virtually on July 16, moderated by Jason Porter.
What is going on?
Some agencies are moving fast. The Inspector General offices in the Department of Health and Human Services (HHS) and the US Postal Service (USPS), for example, are finding that AI is making auditing much easier. Chris Chilbert, Chief Information Officer (CIO) at the Office of Inspector General (OIG) at HHS, said his auditors have used predictive analytics for years, and AI is making it more efficient and effective.
Chris also mentioned that with AI tools, auditors could create and test models more rapidly and accurately. The modeling allows for auditors to quickly process large volumes of data and focus on specific areas of interest. This is a great example of how AI provides for more capacity with more accuracy, elevating the agency’s ability to focus on its mission.
Meanwhile, Gary Barlet, CIO in the OIG at the USPS, said AI is powering sentiment analysis to help inspectors pinpoint problems with mail delivery. This technology also assists USPS with many aspects of its customer service, while allowing more process enhancement, work force utilization and transparency. As a key tool for faster and more accurate mail delivery, AI would help in many instances, such as mail-in ballots for elections.
Gary added that analyzing vast amounts of data might be the single best use of AI right now. Agencies —any kind of organization, really— gather far more data than a human being could ever hope to mine in order to discover connections. AI combs through data to find relationships between data sets that we would never have thought to look for.
In my previous blog, I noted that President Trump issued an executive order in February 2019 titled “Executive Order on Maintaining American Leadership in Artificial Intelligence,” defining AI as a priority in research and development. It is very positive to hear and see government agencies utilize or begin to adopt more AI and machine learning. In many cases, the deployment of such technology can be a game-changer for the agency.
Make the right choice
If you have gotten the impression that AI technology fascinates me, you are right. I do believe it is the future of how technology will be utilized for services and decision-making. However, it may not always be the best option. In my earlier post, I noted that, in many cases, an agency’s current and future needs could be met with a less expensive machine-learning solution.
It actually goes deeper than that. As Microsoft’s Kent Cunningham said during the panel discussion, business outcomes should always drive technology decisions. Agencies do not need to identify an internal AI specialist before starting conversations with a technology provider. However, they do need to define their business needs with precision.
Start with the business need, and then choose the technology that will best address it. Sometimes that will be AI and sometimes it will be something else.
Key questions to ask
As you contemplate an AI solution, consider these questions:
Can you handle the output? If you cannot absorb the product of the AI analysis, you can create additional operational risk. You should focus on change management and your absorptive capabilities, as well as the technology.
Is the outcome defined? Did you specify the requirements? With AI, you can truly refine and be very specific on the information, process or automation you want to achieve.
Do you have the required data? AI and machine learning need data to run the engine. Ensure that you have the amount of data required to process in order to attain your expected outcome. Can you connect the data sources?
Where is your data coming from? AI will process large volumes of data, but if the data is not accurate, the outcome will not be either. Ensure that you have reliable and credible data sources.
To learn more about CGI’s capabilities in AI and related technologies such as machine learning and robotic process automation feel free to contact me, or visit our Intelligent Automation page as a start.