The thought of power-hungry robots replacing humans in the workplace is a major turnoff for people who are already weary of artificial intelligence and what it means for their job security.
Can you blame them? You don’t have to look far for talking robot heads that can hold casual conversations with humans, supercomputers that can beat the smartest Jeopardy players and a growing number of machines that can think for themselves.
It’s a common concern that comes up during AI discussions, along with questions about privacy, security and accuracy of the results that AI produces. Those are issues still being worked out in both the public and private sectors. But in terms of the workforce, proponents of the technology argue that AI is an assistant, not a job replacer. The tech still relies on human smarts, and not every problem can be solved with AI.
For a growing number of federal agencies, AI is being used to quickly tackle mundane tasks and free up employees to solve more complex problems. At a time when many agencies have been forced to downsize their workforces, some leaders have embraced AI to help fill in the gaps.
For example, at the Bureau of Labor Statistics employees process millions of data points every month, including survey responses about workplace injuries and illnesses that are used to identify preventative measures. The agency turned to AI in its quest to find ways of processing data more efficiently, accurately and consistently.
One year, the agency received more than 2,000 distinct job titles to represent janitors or cleaners who have been injured on the job, William Wiatrowski, Deputy Commissioner at BLS, said during a recent event hosted by the Partnership for Public Service and IBM Center for The Business of Government. Trying to slot all those responses into one category was a manual and tedious task. In fact, 80 percent of the surveys never used the words janitor or cleaner.
Despite the growing use of AI to do these and other coding tasks, the technology still has its limitations. “Machine learning couldn’t do everything and, in fact, it was the difficult cases — the ones that they [employees] enjoyed noodling over and finding the right answer to — was still theirs [to solve],” Wiatrowski said.
The National Institutes of Health, more specifically the agency’s grants side, is also using AI. NIH undertakes a massive effort to categorize spending across hundreds of categories. To do this, the agency creates a catalogue of about 500,000 projects and categories each year to report publicly, said Richard Ikeda, Director of NIH’s Office of Research Information Systems.
“To do this manually would take about 700 full-time equivalents, but we do it in an automated fashion with text mining with about 10 percent of the expense of the manual effort that is required.”
Right now, it is a rules-based, text mining system, which means that employees establish the categories in a manual fashion. There are a set of rules that allow research and projects to be properly categorized. But NIH is experimenting with machine learning to train the system to handle these types of tasks, particularly when it comes to categorizing similar projects.
The CIA is also eyeing the benefits of AI. Currently, there are 130 AI and machine learning projects underway that the agency is actively monitoring, said Teresa Smetzer, Director of Digital Futures at the CIA.
The goal is to move from a world where the agency reports on things that have happened, to anticipating, preventing and even and pre-empting events. “It’s still a very early journey,” Smetzer said.
To help agencies that are getting started with AI, experts at the Partnership and IBM offered these tips:
1. Not every task should be augmented by AI, said Claude Yusti, Federal Cognitive Solutions Leader at IBM Global Business Services. Start with tough business problems and consider how AI can help accelerate solutions, but be selective. For people with too much information on their hands who are trying to do their jobs, consider how to use AI as an assistant.
2. Do not underestimate the upfront investment needed. Part of that investment is education, Yusti said. How do you define AI for your organization? AI is on a learning curve and is best suited for small initiatives. Anticipate that some number of the projects won’t go as expected. So take small steps and let AI prove itself.
3. Agency expertise in AI could boost AI’s potential, said Mallory Barg Bulman, Vice President of Research and Evaluation at the Partnership. It always comes down to the people. Make sure the people are managing the change, deciding the work that will happen and shepherding the work.
For more on AI in government, including case studies, check out this joint report by the IBM Center for The Business of Government and the Partnership for Public Service: “The Future Has Begun: Using Artificial Intelligence to Transform Government.”