Artificial intelligence has “enormous potential for government” with the potential for savings in manpower and improvements in service to the public, according to a report from the Partnership for Public Service and the IBM Center for the Business of Government.

“Government leaders working on agency issues, such as how to code data or buy goods and services more efficiently, turn to AI to streamline processes, relieve employees of tedious tasks and provide new insights into their agencies’ work,” it says.

While the term artificial intelligence covers a variety of applications, the report mainly focused on one form, machine learning, in which systems learn from data to improve over time. It cited applications that could be put to wider use including use by: law enforcement agencies to better plan patrol routes including by making them more difficult for criminals to anticipate; by the Bureau of Labor Statistics to relieve employees of tedious tasks such as reading survey responses about workplace injuries, and allow them to concentrate on more important work; and by a state government to identify people at risk of arrest in order to get them into preventive services, in the process breaking down information silos across agencies.

It also cited the Air Force’s plans to use AI to “make sense of complex acquisition regulations so it can speed the process of buying goods and services. Doing so could open government procurement to more small businesses and companies that have avoided working with agencies because the acquisitions process has been too difficult to navigate.”

The agency will “will upload thousands of regulations, contract cases, acquisition training material and Defense Department policy. AI technology then will be able to answer queries from federal contract officials and contractors about acquisition rules and regulations, such as how to proceed with a contract, what procedures to follow or what contract a small business could bid on.”

The report provided this advice for government leaders in adopting artificial intelligence: not every task lends itself to AI; do not underestimate the upfront investment needed; start small; the more and better quality the data, the better its performance; agencies will need in-house expertise in AI but such employees likely will be as difficult to recruit and retain as other cybersecurity experts; and consider working with outside experts, particularly in academia.