Can AI help simplify federal acquisition?

For all the mystery still surrounding the future of artificial intelligence, some early governmental uses could help federal workers and contractors navigate acquisition regulations to make federal purchasing less complex.

A new joint report from the Partnership for Public Service and the IBM Center for the Business of Government pointed to a pilot program currently deployed by the Air Force as a test case for making contracting quicker and more efficient.

Strategic, widespread use of AI "could save government up to 1.2 billion work hours and $41.1 billion annually," the report estimated.

Because of the density and complexity of the current federal acquisition process, "people tend to go for what they know will work rather than where is there opportunity for innovation, saving time and saving money," said Mallory Bulman, the vice president of research and evaluation at the Partnership.

The Air Force, which spent $53 billion on products in services in fiscal year 2017 alone, will be working with two contractors and "hopes to unveil the AI system both online and as a phone application later in 2018," according to the report.

"The goal is to enable a contract officer to query how a specific contract should be structured or if a particular contract type can be used for buying a particular product or service," it stated.

The payoffs of using AI -- in terms of saving both time and human headaches -- could make purchasing easier and more innovative.

Specifically, using AI could allow businesses of any size to query the software for bidding details, as well as a list of contracts they would be eligible to bid. Making contracting easier could lead to a greater deployment of technology and even make bidding easier for small businesses.

The report noted that to prepare the system, contracting officers must first upload massive amounts of data to inform AI's decision-making -- a "major effort before the system can be helpful."

If the Air Force pilot program is successful, it would offer agencies a chance to repurpose the hours that would have been spent on the lengthy acquisition process, as well as "an AI approach to emulate to simplify and speed up what is now a mystifying government necessity."

The report also pointed to three other possible use cases: helping law enforcement, relieving human workers from tedious and rote tasks, as well as helping the country's most vulnerable populations.

However, Federal Cognitive Solutions Leader at IBM's Global Business Services Claude Yusti pointed out that AI in government is still in its nascent stages.

Right now, "AI is Kitty Hawk," he said, advising agencies not to underestimate the amount of upfront investment needed. "AI is really best suited to pick a small initiative around the problem... Pick something that can prove it can pay its way."

Bulman added that any effective AI solution "is always about the data" and agency personnel expertise.

"It really matters what type of information goes into the AI system because that is going to be the basis of the learning," she said.