The Bureau of the Fiscal Service has been using intelligent automation to improve its business operations for the past four years. In that time, the agency has implemented more than 100 bots responsible for saving 186,000 employee hours and about $13.6 million.
Now, the Fiscal Service is ready to dive deeper into automation, looking at the potential of automation and artificial intelligence (AI) to further enhance federal operations.
Though Caitlin Gehring, the Fiscal Service’s acting chief customer officer, said automation is not always the answer — she acknowledged the potential for it to drastically save time and effort when applied to the right processes or tasks.
“We have to make sure that any kinds of automation we introduce is always contained and secure so that we’re not setting ourselves up for failure and causing havoc across the federal environment,” Gehring said onFederal Insights – Intelligent Automation. “So that is a big part of our kind of calculus when it comes to where and when we insert automation.”
Future of automation at the Fiscal Service
When it comes to understanding how automation drives value, Gehring said it’s all about “your customers” and users’ pain points and limitations.”
Gehring said one pilot project at the Fiscal Service is focused on analyzing call center data to identify patterns of user frustration — including things like “mouse thrashes” and “rage clicks” — and helping to improve customer service.
Additionally, the agency is exploring AI solutions for detecting fraud and improper payments. The Do Not Pay service, which screens billions of dollars in federal payments, is currently a manual process. The agency will eventually integrate AI-driven automation into workflows to help prevent improper payments.
Justin Marsico, the Fiscal Service’s chief data officer, said they are also looking at the search function on USAspending.gov, a site that tracks government spending, to see how AI can help with the user experience.
“Instead of having to come and know all the right filters to select and then know how to interpret the results, and to potentially do some additional analysis on their side to get to the answer, [we’d] allow them to essentially type a question in plain language. And then we would translate that into a structured query language, pull the results from the data set, on the back end and then present that to the user in plain text.”
Robotic process automation, APIs
Much of the Fiscal Service’s previous automation success has come with the implementation of robotic process automation (RPA), which has led to significant time and cost savings in the last four years.
Gehring says the 104 bots implemented since 2020 have taken away the need for a human to perform time-consuming manual tasks, allowing them to focus on higher-level, strategic work.
“Given that we handle financial management for various federal agencies, many of our processes were highly repetitive,” Gehring said.
The automation has primarily been deployed in the agency’s Administrative Resource Center (ARC), which provides services like financial management, procurement and human resources for other agencies. These automated solutions have not only improved efficiency, but have also helped reduce employee burnout as well as the potential for errors in high-stakes financial systems.
In addition to RPA, APIs — sets of rules and protocols that allow different software applications to communicate with each other — are also playing a key role in the agency’s automation strategy.
“APIs allow us to automate data exchanges between systems, eliminating manual tasks and reducing errors,” Marsico said. “For example, we’ve automated data publication processes on sites like fiscaldata.treasury.gov, using APIs to pull and share financial data seamlessly.”
Data quality drives automation success
Quality data plays a huge role in supporting the agency’s automation efforts.
“It’s important to make sure that when automation use cases are relying on data, that we understand what the quality of the data is, where it is, how accessible it is, how complete it is and how much we can trust it for decision making,” Marsico said.
The agency has implemented rigorous data validation measures, particularly in important processes like payment and grant distributions. It has also developed automated reporting processes, which provide critical HR analytics to managers, helping them make faster and more informed decisions.
Marsico said his team is also working to fully automate the manual reporting processes within the agency, reducing chances for error and speeding up the data analysis process altogether.
“We want to make sure that we are giving back information that’s actually accurate and that the users can believe in,” he said.