Your browser is unsupported

We recommend using the latest version of IE11, Edge, Chrome, Firefox or Safari.

A Conversation With…

The Government Finance Research Center works with researchers from a variety of backgrounds to analyze the role that public finance plays in our lives. In the interviews below, we talk with experts to dig deeper into pertinent topics and get their perspective on the past, present, and future of government finance.

Chris Atkins, Vice President for Digital Government Transformation, SAP and formerly chief financial officer, Indiana Heading link

photo of Chris Atkins

Q. I know you’ve given a great deal of thought to the use of artificial intelligence – AI – in public sector finance. Could we start out by defining AI?

A. That’s a tricky question. But I’d say that AI is the use of technology to enhance your ability to understand your data and to automate processes using that enhanced knowledge. It covers a lot of different things. It’s machine learning. It’s robotic process automation. It’s a suite of intelligent technologies that allow you to improve existing processes or extend existing processes into new areas.

Q. Bearing that in mind, can you comment a bit on the utility of AI for the public finance community?

A. It helps to eliminate a number of processes that have required a lot of manual effort on the part of government workers — things like processing invoices, processing job applications, auditing. AI because it’s flexible can deal with lots of different potential scenarios and anomalies, unlike traditional ERP systems that regulate processes that are virtually the same in every case.

Q. You mention processing invoices. How can AI help deal with something that seems so mundane?

Sure. You need to match up the invoice to the payee, which could change year to year. And that can require a lot of manual effort. But tools like AI can take irregular processes like that and look for patterns and help integrate those different patterns into your standard financial processes. Also, with AI, finance officials can look at those patterns and say, ‘Hey, the last time this happened, something else happened.’ For example, the last time sales tax submissions dropped to this point, we had a recession and we had to cut spending by a certain amount. It can help to alert you to things that are happening so you can get ahead of them.

Q. That’s clear. Can you give us a real-world example of the utility of AI?

Pennsylvania, for example, felt their internal audit efforts were way too manual and they weren’t well staffed and they couldn’t do what they really wanted to do in terms of better targeting those activities they felt were likely out of compliance in terms of spending.

In the past, they would audit a lot of transactions that were perfectly in compliance, which was time consuming. But if they looked at the patterns of spending or transactions from the past, they could find those that were likely to be out of compliance and focus their efforts there. Our partner The Solutions Company used the SAP Business Technology Platform to develop an algorithm for them that can look at historical patterns and then identify those patterns in their current year spending and then say, ‘Hey, you should go focus your audit efforts here.’

Q. It would seem like AI would be helpful in detecting inefficiencies or even fraud.

A. That is very accurate. CFOs will often put out guidance about internal financial management practices, like procurement cards or travel. So, if state employees are all going to the same event, they should all travel together. Four people shouldn’t generally use four different fleet vehicles if they are going to the same place. But it’s tough to track that with many existing systems. Looking at data with AI, you can see that these four employees went to exactly the same place and all of them checked out a car on this particular date.

Q. That’s a great example. Another one?

Yes, one of our customers is Office of State Revenue in New South Wales, Australia, which is in charge of ensuring that once a taxpayer owes a debt then they are going to collect that debt. So, we looked at their historic collections data. And what we found was that a first letter would be sent out and there’d be no response. A second letter would get no response. And so on until the final letter was sent. And then there was a group of taxpayers, who would pay almost immediately after they got a letter that said this was the final one before they’d be taken to court.

So, they started to just send one letter, the final letter, and that accelerated their collections. And now they’re able to collect money earlier than before.

Q. That’s impressive. But what about people who simply couldn’t pay?

We were able to use the data to learn more about the circumstances that person might be in. Maybe that person had just lost a job or had lost their house because of a natural disaster. We have other data that can tell us this, so instead of taking them through this multi-step process the office would say ‘Hey you owe this debt. And we know that you just lost your job. So, we’re going to offer you a payment plan up front, because we know that you probably need one based on your current circumstances.

Q. What are the challenges to getting the most out of AI?

One of the biggest challenges is the data itself. Data quality is inconsistent from state to state or locality to locality. But in order to use AI you need clean data and not every government entity in the United States has really invested in managing their data as the asset that it is. Finance officers should welcome Chief Data Officers into the government C-suite because better, cleaner data will enable all these finance examples we discussed today.

Another big challenge is just getting access to the data. Government can be very siloed and the people in those silos can be very territorial about many things including giving other departments access to their data. You have to have a framework by which agencies or department can share data or allow another entity that’ working on a project to access the data it needs.

Interview conducted by Richard Greene, senior advisor, Government Finance Research Center

Read More Conversations Heading link