GovLoop

The Human Piece of the AI Puzzle

Agencies will struggle to adopt artificial intelligence (AI) if they don’t consider how it will impact their employees, according to two top federal IT officials.

The pair added that AI – or machines imitating human cognition – will integrate with governments most seamlessly when they consider how it will change their workflows.

“You have to think of the whole system that would be impacted from the beginning,” Dorothy Aronson said Tuesday. “If, in the end, people are left behind or feel threatened, they won’t help. They won’t go to the future if they want to stay in the present.”

Aronson is the National Science Foundation’s (NSF) Chief Information Officer (CIO), and she also serves as Co-Chair of the Federal CIO Council’s Workforce Committee.

Gary Barlet, the CIO of the U.S. Postal Service’s (USPS) Office of the Inspector General (OIG), made remarks about AI that were similar to Aronson’s Tuesday.

“There are lots of foundational things you have to put in place,” he said. “There’s the matter of selling it to executives. If you jump too far ahead, you’re going to fail.”

Aronson and Barlet were both speaking at the Digital Government Institute’s (DGI) and 930Gov’s IT modernization conference in Washington, D.C.

According to Aronson, NSF recently tried using AI to help the agency’s program managers pick speakers with the right educational backgrounds for events.

Aronson said that although the AI was useful, it also caused concern among NSF’s workforce about their job security.

“What happens if there are people whose whole job it is to find people for these panels?” she asked. “Solving the problem of what to do with the people who were left behind was very difficult.”

Aronson said that NSF has about 2,000 employees, with roughly 200 of them using the agency’s speaker tool before AI entered the picture. Since AI’s appearance, approximately 50 people have lost some of their work to the new technology.

“It didn’t eliminate any jobs entirely,” she said. “It merely freed up some time. It’s learning how to look at your work, so you can then discard it. People don’t necessarily look at their workload and think, ‘This is something I can automate.’”

Automation involves machines performing performances and procedures with minimal human involvement.

Barlet said that USPS is examining whether AI can use automation to assist its OIG with investigating surveillance videos.

“They record weeks of video,” he said of OIG’s investigators. “The lucky investigator gets to sit down and fast-forward for what they’re looking for. There’s a lot of dead space in that video.”

Barlet added that AI could speed up OIG investigations by looking for suspicious activities in the agency’s surveillance videos without humans.

“It will free that investigator up to go and do other things instead of sitting at that desk watching that video,” he said.

According to Barlet, USPS’ OIG has about 500 investigators and 300 auditors. AI would save these employees scores of hours on trivial, manual tasks such as video surveillance or filing paperwork.

For example, Barlet continued, USPS OIG investigators have long had to manually scan the agency’s reports of investigation (ROI) to send them digitally to U.S. Attorney’s offices.

“Anyone who has scanned anything knows that you’re one paper jam away from restarting,” he said. “We’re talking about getting the users brought in. Find something they hate and automate it. Let them focus on the things that they enjoy doing.”

Ultimately, agencies that implement AI can use it to automate manual, repetitive work so that their workers can focus on more satisfying efforts. Over time, these employees become more satisfied while saving their energy and time for activities they enjoy.

Aronson noted, however, that many agencies must factor their people into simpler changes such as IT modernization before tackling AI.

“IT is not the place to start,” she said. “The place to start is with the problem. What problem do you want to solve? It’s people first. You apply IT if it will help you get through the problem.”

Exit mobile version