The use of AI applications in government looks to be getting a lift from a new policy focus and the perennial need to improve citizen services with fewer resources.
The upshot could be more AI advisory and implementation work for IT consultants focusing on the government space. Consulting firms and market researchers presenting at this week's AI World Government conference in Washington, D.C., pointed to the field of conversational AI as a particularly promising area.
Impetus for greater AI adoption comes, in part, from the AI in Government Act, which aims to promote agency use of AI for the public benefit. The Senate bill (S. 3502) debuted in September 2018, while the House introduced companion legislation in May (H.R. 2575). The House bill calls for the creation of an AI Center of Excellence to advise agencies on "developing innovative uses of artificial intelligence."
"This is a hot area," said William Meisel, president of TMA Associates, a consulting firm that focuses on natural language processing applications. "Government agencies are going to be asked about doing something about AI."
Meisel, who chaired a seminar on conversational interfaces at AI World Government, suggested agencies face public pressure as well. Market research bears this out: A 2016 Accenture survey of 3,000 U.S. citizens revealed 85% of the respondents expected "the same or higher quality from government digital services as they do from commercial organizations." And in 2019, 86% of the respondents to an Accenture poll of citizens from the U.S., Australia, France, Germany, Singapore and the United Kingdom said they view digital delivery of public services "as equally or more important to them than traditional methods."
Conversational AI in government
Agencies looking to deploy AI-based digital services may find conversational AI a starting point. Some agencies have already deployed intelligent virtual assistants (IVAs) or chatbots, with the Army's SGT Star and the Citizenship and Immigration Services' Emma virtual assistant among the more prominent examples. Other public entities following their path may find a fairly rapid ROI and lower risk compared with other types of AI applications in government.
Ron Schmelzer, managing partner and principal analyst at Cognilytica, a Washington, D.C., market research firm, said conversational AI systems have "great short-term ROI," providing immediate benefits in use cases such as password resets.
Ian Beaver Lead research engineer, intelligent self service, at Verint
Agencies can start small with password resets and then let the conversational system handle more complex activities, added Kathleen Walch, managing partner and principal analyst at Cognilytica.
In contrast, other AI patterns involve a more extended ROI timeline. Schmelzer cited predictive analytics, which takes longer to validate, and autonomous systems, which are more difficult to implement.
Meisel, meanwhile, said conversational interfaces for customer service offer a low-risk use of AI. Customers are already frustrated with touch-tone menus and being placed on hold with traditional customer service approaches. So the potential to improve the customer experience is significant. In addition, such systems can be equipped with a safety net, in that human agents can take over a customer's inquiry if automation fails, Meisel explained.
Motivations for adoption
Ian Beaver, lead research engineer, intelligent self service, at Verint, a Melville, N.Y., company that provides customer engagement consulting services, said natural language self-service systems can help government agencies cut costs, scale to meet demand and provide insights into customers.
He said prospective customers sometimes push back on the need for natural language self-service, asking, "'Isn't directed IVR, [or interactive voice response], an FAQ on our website enough?' I would argue that it is not."
IVR, FAQs and knowledge bases are good resources to employ, Beaver said, noting he would not recommend eliminating them. But conversational AI provides the additional benefit of open-ended customer feedback. That feedback can uncover opportunities to improve other aspects of citizen outreach, such as public-facing websites. For example, when Verint helps a client with a natural language system, the company may find users are telling the IVA that they can't find certain information on the client's website.
Agencies that forgo natural language customer service miss out on this benefit, Beaver said.
"This sort of feedback mechanism is only possible if you let the human tell you what they are after," he said. "With directed IVR and directed dialog, … you are predetermining the types of interactions that your users can have with you. You are deciding what they say. You are deciding the path they can go down."