Chatbots vs. RPA vs. virtual assistants
Improving customer service is an imperative for governments at all levels as evidenced by the goals of the President’s Management Agenda and the National Association of State Chief Information Officer’s top 10 priorities list for 2019. To do it, more agencies are turning to emerging tools such as chatbots, virtual assistants and robotic process automation (RPA).
So what's the difference and how are agencies using the new technologies?
Chatbots are among the most common of the three applications, although the three are interconnected. “Short for ‘chat robot,’ a chatbot is a computer program that simulates human conversation, or chat, through artificial intelligence,” according to DigitalGov.gov. Conducting conversations via auditory or textual methods, chatbots ultimately answer users’ questions or link them to the people or resources that can.
Although building a chatbot takes considerable work -- cleaning, metadata-tagging and organizing data and tying it to natural language processing and other technology -- the benefits of chatbots can be substantial. The include the integration and management vast amounts of data from multiple sources, identification of patterns based on customer interactions and reduction of the amount of time employees spend on “mundane tasks,” Adelaide O’Brien, research director at IDC Government Insights, wrote in an email to GCN. The result is better served citizens and more engaged employees.
“Bots can provide 24 x 7 customer assistance. Bots can ask questions about individuals’ needs, habits, and preferences and offer contextual and personalized services based on a combination of constituent supplied data and data from other sources,” O’Brien said. “Chatbots will significantly increase the capability to process vast amounts of data and reduce the need for employees to perform repetitive tasks, such as [answering] questions regarding services, qualification for benefits, and eligibility for refunds, freeing employees to engage in more difficult decisioning.”
She pointed to a bot that the NASA Shared Services Center at the Stennis Space Center in Mississippi uses to streamline workloads of employees including the space agency’s CIO and CFO. In May 2017, NSSC, a fee-for-service organization that provides agencywide human resources, procurement, financial management, data center support and service desk transactional services, launched George Washington, the federal government’s first digital bot, to help with human resources and procurement processes.
“Washington bot, with its email account, credentials to access operational systems, and ability to run based on receipt of emails, was integrated with NSSC staff,” O’Brien said. “Whether Washington ingests structured emails to create tickets for new employee suitability adjudication or logs into the financial management software to enter human-approved budget information, it acts and looks to employees like any other employee that reads and processes emails, looks for new files in group folders, provides feedback in real time when exceptions to work instructions are encountered, and reports all of its work upon completion.”
In June 2017, NSSC added the John Adams bot for finance-related work. Today, it also has the Pioneer bot, which creates procurement requests for the Office of the CIO, and plans to put a fourth bot, Beacon, into production in early 2019.
Talking up chatbots
“We’re pretty early on in this market because only about 10 percent of enterprises have production chatbots in deployment, based on some research we did a few months ago,” said Van Baker, a research vice president for Gartner’s Application Innovation team. But “I’m getting lots of questions from federal, state and local governments around the deployment of chatbots to basically improve accessibility to the services by the citizenry that they serve.”
For instance, U.S. Citizenship and Immigration Services’ Emma is a bilingual interactive assistant that sits in the top right of the agency’s website waiting for users to click and ask questions. Another is the Army’s Sgt. Star, which can answer questions about joining the service, including pay rates and enrollment processes.
Since launching a chatbot in April 2018, the Montana Department of Justice’s Gambling Control Division has seen a 47.31 percent interaction rate with the bot. That has reduced the number of phone calls that workers need to handle, Anne Gerken, communications specialist at the division, wrote in an email to GCN.
“Feedback from our Investigators who interact with licensees on a regular basis is that they really love the fact that they can open it up on their cell phones and either access a concrete answer or provide a quick tutorial on how to use the bot instead of having to search through our website,” Gerken said.
The division updates the bot based on the questions it receives. “If we notice a new pattern, we add to our Frequently Asked Questions database (which is entered into the bot),” she said. “We might also update common terms used for searching the bot." The office in the process of creating a new bot that will have more capabilities for directing questions, she said. "Ideally it will be more useful for patrons by providing more specific direction in their single-word searches.”
Eventually, Gerken said, the division wants to use the bot to help licensees find reporting forms, applications and statistics -- and perhaps even submit applications.
A key benefit of chatbot’s conversational user interface is that it’s easy to use for both internal and external users. “Everybody knows how to ask a question. There’s no learning curve associated with it if you do a good conversational user interface,” Gartner’s Baker said. “Once the question or the request is fully understood by the chatbot or virtual assistant, that can kick off business processes that run in the backend.”
Mississippi Interactive (MSI), a subsidiary of NIC, is looking into ways to make transactions happen within the state’s chatbot, Missi, said Dana Wilson, MSI’s general manager. For example, users would be “able to ask Missi when your driver’s license expires and then actually have her walk you through the process of completing that transaction would be a future goal of where we see chatbots going,” Wilson said.
Since launching Missi in 2017, the bot has taken about 24,000 questions -- a number Wilson said has surprised officials. “We don’t get that many phone calls, so it’s interesting to see that many people taking advantage of an artificial intelligence platform specifically for government,” she said. “You can equate that to being phone calls or emails, [and] you could really see the value of an artificial chatbot being able to field those questions any time of day as opposed to taking those calls and having a human interaction, when they can dedicate those resources to more pressing items.”
MSI also studied analytics around the top search terms and website traffic to determine what information to offer first.
“We looked to our site analytics, took the top searches, and those were the first things to onboard to the chatbot since those were the majority of what people were coming to look for online,” said Drew Levanway, MSI’s director of operations.
In Los Angeles, the Information Technology Agency saw results quickly when it launched its Chip chatbot on the Business Assistance Virtual Network in March 2016. “We saw an immediate drop – almost 70 percent – of the general email questionnaires that were being asked, so that was a cost-savings for my developers’ time,” said Joyce Edson, LA’s deputy chief information officer. Chip, short for City Hall Internet Personality, has had 20,000 conversations and fielded about 100,000 questions.
Now, ITA is working on LACIE, which stands for the Los Angeles City Interactive Experience, which Edson expects to come online in early 2019. “We want to bring her to life with location-based [augmented reality], kind of in the vein of Pokemon” Go, she said, to move bots from a disjointed voice through the Amazon Echo and Dot, Google Home and Google Assistant and Chip to a virtual physical presence.
The potential for chatbots to use RPA, the automation of repetitive customer service tasks, to automate processes that are currently manual, such as renewing licenses and registering properties, is real, Gartner’s Baker added. If a chatbot can “walk somebody through the steps that they need … that’s going to both benefit the citizenry in that it makes it easier for them to do," he said. "It also benefits the government in that their facilities are in many cases overtaxed and just don’t have the resources and the personnel available to be able to deliver the services in a timely fashion, so it serves two purposes,” he said.
RPA also increases accuracy, speed and standardization and can save between 40 to 70 percent on labor costs, with near zero error rates in both front- and back-office functions, according to a recent NASCIO report. It can also be used across multiple systems.
The NSSC bots use RPA, for example, as do the Health and Human Services Department’s HHS Accelerate, which uses it for acquisitions, and the Defense Information Systems Agency, which pilot tested bots that automated aspects of the auditing process in its Accounting and Readiness Division.
The General Services Administration has looked at using RPA to “open our programs to better decision making through AI,” Keith Nakasone, deputy assistant commissioner for acquisition operations at the agency’s Office of Information Technology Category, told the Oversight and Government Reform Committee’s IT Subcommittee on March 7, 2018.
GSA has six active bots, with four in the CFO's office and two in the Public Buildings Service. One application enters information into GSA's financial system for purchase card holders, and another sends notifications when invoices are nearing their due date. In 2019 the agency plans to deploy RPA on larger, "higher value" applications such as data entry, which will save automate hundreds of thousands of hours of human labor, GSA's RPA Program Director Ed Burrows said late last year.
On the same continuum as chatbots are virtual assistants, Baker said. Gartner defines these tools as “a conversational, computer-generated character that simulates a conversation to deliver voice- or text-based information to a user via a Web, kiosk or mobile interface.” A virtual assistant takes chatbots a step further, he said, because the technology could involve many chatbots and use cases and automate several functional areas. A virtual assistant can capture a user’s behavior and adapt how it interacts with that person based on activity patterns.
Another difference is that unlike chatbots, virtual assistants are usually dedicated to managing a few users' needs, said David Schubmehl, research director for IDC’s Cognitive/AI Systems and Content Analytics research. “For example, having a virtual assistant [could help] me with my calendar scheduling and/or being able to answer questions and potentially being able to surface content related to things I’m working on,” he said.
One of the more well-known virtual assistants is Amazon’s Alexa. The voice-enabled system can set alarms, make to-do lists, play music and radio news and control smart-home devices. Users can download additional "skills," or third-party apps, to extend Alexa's abilities.Mississippi became the first state to launch an Alexa skill in 2016. It started with frequently asked questions and has since expanded to include responses to queries such as “Who is my elected official?” or “What is the number to reach the Public Safety Department?” MSI’s Wilson said. It has since expanded to include real-time traffic alerts and news from Mississippi Public Broadcasting, and in 2018, the state made this information available to Google Home users as well.
“We started with the Alexa skill because of the development resources that Amazon specifically had available, and so it was a quicker turnaround,” Wilson said.
The call for IT modernization isn’t only about replacing and revamping aging equipment. It’s also about implementing cutting-edge technology such as chatbots, virtual assistants and RPA, and the clamor for more of these is only getting louder.
An Aug. 27, 2018, memo from Office of Management and Budget Director Mick Mulvaney stated that the agency plans to use RPA and AI “that will execute repetitive administrative tasks and significantly reduce the burden on Federal employees.”
GSA's Emerging Citizen Technology Office has an open source pilot that has helped federal agencies make their services available via Alexa in addition to her counterparts Google Assistant and Microsoft Cortana.
At the state level, the number of CIOs who said AI and machine learning are the most impactful emerging IT in the next three to five years rose to 58 percent from 29 percent between 2017 and 2018, according to a NASCIO report.
Gartner’s Baker says the future is bright for chatbots and virtual assistants: “It’s a very compelling value proposition for the user because you don’t have to learn anything. All you have to do is be able to ask a question.”