GAO imagines the future of oversight

Auditors at the Government Accountability Office typically evaluate the success or failure of government programs retrospectively. But at the GAO's new Innovation Lab, there are efforts afoot to shift the way oversight is done to provide real-time and even predictive analytics about program performance.

The new Innovation Lab launched by at the GAO's Science, Technology Assessment, and Analytics team is designed as a kind of audit and oversight sandbox to test efforts to use predictive analytics, large data sets and artificial intelligence to spot trends and anomalies in government programs – especially with regard to spending. Fraud and improper payments have long dogged some of the big benefits programs and aroused anger in Congress. The lab is looking at the possibility of actually doing data mining of entire government datasets rather than relying on statistical sampling.

"Imagine, for example, being able to throw the analytical power of AI and machine learning at the billions of data points in the Department of Treasury’s general fund and getting to the point where you could identify every improper payment without the need for any statistical sampling at all," the agency said in a blog post.

Taka Ariga, GAO's chief data scientist and director of the Innovation Lab, is also bullish on what the lab will do for the oversight agency and the field of auditing at large.

Some challenges there include developing ethics for the use of AI and algorithmic analysis in auditing, and methodology for preserving security and private in datasets.

"We're seeing a signal for GAO leadership in that area," Ariga told FCW. "GAO is uniquely positioned to serve that role."

The lab is looking to improve and scale technologies to help GAO operate more efficiently. Ariga said the lab is looking at robotic process automation to help GAO manage internal processes as well as improving GAO capabilities in handling data sets so that the subject matter experts can work on data without a lot of time consuming, labor intensive prepositioning activities.

As the lab develops prototypes and pilots, Ariga said, it is looking to potentially share with the audit community via a GitHub repository. Challenges remain, however, in terms of sharing code without compromising the sensitivity and privacy of the underlying data.

Ariga has worked around government in top professional services firms including Deloitte, Ernst and Young and Booz Allen. GAO is his first stop at a government agency – he joined four weeks ago and has largely been tending to "startup activities" for the lab to get the new unit operational. He's embarked on an "internal roadshow" at GAO to understand challenges that are facing audit teams and how the new lab can help.