Whether they call them chatbots, digital assistants or conversational artificial intelligence (AI), companies are using technology to enhance HR service, boost productivity, and help deliver more rewarding and personalized experiences for job candidates and employees.
"One of the most visible applications of artificial intelligence are conversational agents—chatbots and intelligent assistants that interact with people via voice or text channels," said Arthur Franke, a data scientist and director of data and analytics at KPMG in New York City. "Having conversations with AI is becoming routine for consumers, and soon it will be for employees, too."
Chatbots will provide value to HR in a number of ways, experts agree. "In the past, an employee with a question about how to get something done would ask a knowledgeable colleague for an answer," said Franke, author of a KPMG report on conversational AI. "In the future, they'll ask a conversational agent, and artificial intelligence will answer their question."
It's already happening, as employers have begun deploying the technology for HR purposes. "AI and conversational chatbots in particular are useful in enhancing and automating the experience of repeatable tasks," said Jeff Mike, vice president and head of research ideation for Bersin, Deloitte Consulting. In HR self-service centers, bots are automating high-volume tasks such as changing an address or updating benefits information, he said. On the talent acquisition side, bots can deliver "a streamlined candidate experience for high-volume recruiting activities" and guide new hires through the onboarding process, Mike said.
"The use of chatbots has exploded in talent acquisition in recent years," said Ben Eubanks, SHRM-SCP, principal analyst at Lighthouse Research & Advisory, a human capital research and advisory services firm, and author of Artificial Intelligence for HR (Kogan Page, 2018). "Recruiters don't have unlimited time to monitor the careers site and answer every applicant's question. Bots are handling initial screenings, scheduling interviews, answering candidate FAQs and even checking in with new hires to make sure they're settled. Recruiters' time is freed up for more high-value work, like sourcing hard-to-fill roles or negotiating an offer with a candidate on the fence."
One company Eubanks worked with said 80 percent of employee service requests that came into HR were handled by a bot within a month of the technology's launch. "Other companies I'm talking to say that using a bot saves their recruiters seven to eight hours each per week, mostly in early screening and interview scheduling," he said.
What Makes a Good Bot?
Franke explained that conversational AI leverages machine learning algorithms to extract information from large datasets. "When conversational agents are built properly, end users do not have to speak any differently to the machine than they would to a human," he said. "The answers that conversational agents give are appropriate—and ideally, useful, constructive and accurate—even when new situations present themselves."
It's important that chatbots focus on repeatable tasks and also not lose sight of the human user, Mike said. "You want the experience to be as conversational as possible, and not mechanical. The algorithms and the data used to train the bots need to be continuously monitored and audited against bias. The bots also need to be able to continuously learn about individual users to enhance personalization."
Most bots today don't explicitly learn on their own, Eubanks noted. "Everyone talks about the self-learning bot, but bots are essentially a conversational interface for a database. You query something, and it responds back, similar to Googling something and getting a result."
Eubanks explained that when chatbots don't know how to answer something, the questions are typically routed to a human "co-pilot." More sophisticated bots can watch the human type the answer in and pick up on it for the next time it sees a related query, but most bots on the market today would need to have that information programmed in, Eubanks said. Either way, failure should be a learning opportunity.
Being able to personalize conversations is really important, Eubanks said. "A chatbot can have a conversation with thousands of people at once and make each of them feel like they are connected and having a real discussion. And 75 percent of people who end a discussion with a chatbot say 'thank you' to the bot, even when they know it's a bot."
He added that bots must be easy to use and enabled for mobile devices and popular workflow tools like text, e-mail and Slack. "Don't make people go to a separate platform and log in—it's much less likely people will use it."
Adoption Challenges
KPMG found that lack of data from which the bot can learn, compliance risks and accessibility issues are some of the most prevalent chatbot-adoption obstacles that companies face. "The company should use inclusive design practices to ensure the bot interface is accessible to all potential users, regardless of working situation, capability or language proficiency," Franke said.
"The biggest adoption challenge is a lack of understanding of how AI and chatbots work," Mike said. "That means recognizing that the chatbot is not making the decision, but it is informing the decisions made by people. HR needs to get better at understanding how to use data to make decisions."
Tips for Success
Implementing bots in the workplace is a major change and rarely without opponents, Franke said. "There are those who think technology can't get the job done and those who worry it will eliminate their role. Change management and governance are crucial but often overlooked areas of implementing conversational agents."
Eubanks said that organizations should be transparent about using a bot. "People may get irritated if they think they have been talking to a person and end up with an error message." He also recommended that HR keep track of employee and candidate questions to better train the bot, especially if the queries are uncommon.
Mike advised starting small with bot implementation, experimenting and learning and then scaling up. "HR is known as the risk manager, and the nature of chatbots and AI is that there is learning involved," he said. "The notion that everything must be perfect and understood before rolling it out hurts innovation. The organizations moving forward are willing to take calculated risks with technology before expanding its use."