How the pandemic transformed citizen service

COVID-19 put unprecedented demands on government services. Call centers were flooded with questions, new programs launched, in-person channels disappeared, and agencies were swamped with citizen inquiries that crossed organizational silos -- all while agencies had to reinvent their internal operations.

In May, a group of federal IT leaders explored how agencies have responded to the changes and where there may be new opportunities to break through old barriers to deliver better citizens services.

The discussion was on the record but not for individual attribution (see sidebar for full list of participants), and the quotes have been edited for length and clarity. Here's what the group had to say.

Supporting the surge

Not every agency saw dramatic increases in citizen requests. For one agency, call volume dropped by two-thirds, forcing managers to reallocate staff during the lull while remaining ready for demand to bounce back.

For agencies directly involved in relief efforts, however, the work increased exponentially. Traffic to some federal websites spiked 1,000% as citizens sought answers and assistance. One participant told of processing as many transactions in a 24-hour period as that program normally handled in an entire year.

Simply adding more resources wasn't enough. At one agency's call center, the volume "exploded through the roof," one official recalled. "We added thousands of new people in there, but that still couldn't be managed from a human standpoint. The velocity of incoming calls was far exceeding any capabilities to scale up there."

"We had to look for other ways to minimize or reduce the traffic coming in through our customer support channels," the official continued. To ease pressure on the traditional phone channel, the agency created self-service help resources and added chatbots to answer rudimentary questions.

When email inquiries similarly surged, agencies turned to automation and natural language processing. Not only could simple requests be handled without a human at the keyboard, but the analysis provided new insights into what citizens really needed. "We were able to see some trends, some sentiment analysis, to see the pervasive theme that's coming across these millions of emails that are coming in, the official said. "We used that information to then inform our FAQs, policies and inform our program offices and say, 'Look, this is what we are hearing in an aggregated fashion from the customers and the citizens who are the recipients of these services.'"

"Quite differently, we are now listening to the voice of the customer," another official said. "The voice of the customer does not come through the standard channels of filling out our customer survey form. It happens organically. And guess what, we as consumers of technology today have many options of sharing our feedback and the most common of those are in social media platforms. If you're not listening to that, obviously, you're missing the voice of the customer or the citizen here."

A tipping point for natural language processing?

The sheer volume of citizen requests may have finally pushed natural language solutions into the realm of the practical for many agencies, another participant said.

"Commercial partners and large companies have now been embracing natural language, voice assistant machine learning tools for coming up on 10 years now," the official said, "where it's really been a struggle for government agencies until the last year."

The official pointed to two problems. First, "the cost efficiencies really only come into play when you're dealing with extremely large-scale volumes. It makes complete sense for Amazon. It makes complete sense for Comcast. Whether it makes sense for every agency or every part of an agency? Less so."

Second, commercial firms often "try to be so conversational that you lose the certainty of the message that you're delivering. I often joke that in a lot of cases in the contact center space, when acting as the voice of the federal government, we often want our people to act like robots. Why on earth do we want our robots to act like people? We want them to deliver extremely certain phrases. How do we help our consumers fit into the bureaucratic need that we have to get them to the exact right spot? That takes certainty."

COVID drove the volume to create cost efficiencies, the official said, and the natural language algorithms have advanced to the point where they can "understand the intent of the caller," and then deliver carefully "scripted language on the back end to provide an answer. We're finally seeing that technology be available and be effective."

Several executives argued that citizen expectations had changed significantly as the number of requests exploded. "A slightly different aspect of this conversation is the changing nature of what's acceptable," one said. "I think the aspect of how we identify the appropriate ways in a changing landscape to authenticate engagements with taxpayers and the public" is essential.

A longer version of this article was first posted to FCW, a sibling site to GCN.