Frequently Automated Questions: Is Artificial Intelligence the Future of FAQ?

Willyam Bradberry/

Finely tuned chatbots could tackle the public's millions of questions.

In the near future, artificially intelligent chatbots may digitally engage citizens on behalf of the federal government.

Drawing on alternative rock band REM’s album “Automatic for the People,” Justin Herman, digital communities and open government lead at the General Services Administration, recently explained an early business case for AI in customer service.

No, these AI-laced chatbots won’t destroy you at chess or own you at “Jeopardy!” Rather, fine-tuned through machine learning, chatbots could automate some aspects of customer service. The chatbots could respond in real time to millions of questions and comments citizens make through a growing number of third-party platforms and provide federal agencies more accurate data about what services and information people actually want. (ICF has a great visualization illustrating the extent to which federal agencies use verified third-party platforms, such as Facebook, Google, Socrata and Github, to engage citizens).

» Get the best federal technology news and ideas delivered right to your inbox. Sign up here.

“If an agency gets 100,000 questions in a certain period, how many of those can be automated or responded to?” Herman said last week at a Digital Government Institute event.

Most of those questions distill to a few standard frequently asked questions, and even though most agencies have websites dedicated to FAQ, “that’s requiring citizens to go and search and find a website and go on a mission," he said.

“That’s one of the ways where AI for customer service comes in,” Herman added.

Chatbots equipped with sophisticated natural language processing could respond to citizens’ questions, generate reports on what citizens are asking about, and determine what results in a successful transaction or service performed.

A service like that could pay dividends for an agency like the Internal Revenue Service, which was able to answer only 37 percent of the calls citizens placed to it during the 2015 tax season. Calls answered still experienced wait times of 23 minutes, which is an eternity for customers used to banking, shopping and swiping right on their dating apps in a matter of seconds.

“What we have to do is move that transaction out from citizens having to hunt for it, and finding it naturally where it is and even automating it as much as possible,” Herman said.

In June, GSA hosted an AI-related event Herman called “frustrating,” in part because AI is a new technology and expectations for it vary across the board.

GSA’s exploration around early AI use cases mirrors a larger White House effort to better understand the intersection of AI and government. At a Nextgov event last week, Lynne Parker, the National Science Foundation’s director of Information and Intelligent Systems division, said the White House’s summer solicitation for feedback on AI resulted in some 400 pages of comments. Parker said they range from bleak (AI will eliminate jobs and pose a substantial risk to humanity) to optimistic (it'll deliver us a utopia to live in).

In any case, those comments will help shape an AI framework for government and direct federal research efforts. Similarly, early use cases of AI and machine learning—like the ongoing efforts at GSA—are likely to play a major role in how the federal government responds to this evolving technology, too.

“This is far from being a fad,” Herman said. “This is an inevitability.”