Thanks to NLP, chatbots get chatty

Better natural language processing allows chatbots to converse in a more “human” manner. That’s good for business

NLP chatbots
  • The latest chatbots can follow digressions, understand semantic irregularities, and grasp nuance
  • Chatbots will soon handle 85% of customer interactions with businesses
  • Companies are using chatbots to engage with followers and interview job candidates

Considering he’s been dead for 64 years, Albert Einstein is awfully chatty. Or at least his bot is. In 2017, National Geographic unveiled an Einstein chatbot on Facebook to promote its TV series Genius. Users could quiz the late physicist about the theory of relativity, the internet, pop bands—anything—and Albert would respond with deep thoughts and witty quips. Asked “How are things?” the Einstein bot would answer, “Since the universe is expanding, farther apart than when you asked that question.”

The bot was a hit. On average, users chatted with it for 6 to 8 minutes a pop. More than half of them returned for a second conversation.

The Einstein bot was pure entertainment, but it had a quality that is driving more advanced development of AI‑powered chat: the ability to converse with human‑like nuance. Computer programs that can realistically simulate human speech rely on advances in natural language processing and deep learning. Armed with this technology, the latest chatbots can follow digressions, understand semantic irregularities and grasp the emotional tenor of an interaction.

Ger personalized analysis for potential value added

“Even just a few years ago, chatbots were very clunky,” notes Brian Westfall, senior content analyst at Capterra, a service that rates enterprise software. “If there was any minor hitch in your communication, the chatbot would just break.”

The ability to carry on more natural‑sounding conversations means that chatbots can take the place of humans in more and more computer‑based interactions. They can act as a kind of personal shopper, assisting consumers and boosting sales for retailers, online and off.

Computer‑generated recruiters can aid corporate human‑resources departments by guiding job candidates through the application process. Healthcare chatbots can collect medical symptoms, recommend remedies and make referrals to a live doctor.

Consumers increasingly are communicating with brands via these virtual representatives. More than 1 in 3 say they interact with a chatbot at least once a week, according to a recent Forrester study. By 2022, bots are expected to handle 85% of customer interactions with businesses.

Many consumers cite the speed and convenience of talking to a bot, yet when given a choice, most still prefer to talk to a person who can better understand their needs. That distinction may be quickly fading. Tomorrow’s chatbots will be able to engage with users on a more personal level, display unique personalities, and interact with a greater degree of social intimacy and apparent autonomy.

The Einstein bot, for instance, could play the role of the wisecracking friend and even initiate conversations. “Out of the blue, you would get a text from Albert Einstein on a Tuesday night, right before the show, saying ‘Hey, how’s it going?’” says Layne Harris, vice president of innovation technology at digital marketing agency 360i, which created the Einstein bot.

Here’s a look at three sectors where conversational bots are expected to play a big role.

NLP Chatbot Use Cases


In a 2017 survey by Segment, a customer‑data software company, more than two‑thirds of shoppers complained of being frustrated by impersonal interactions with companies. But as retailers increasingly rely on chatbots, virtual agents with a knack for human‑like conversation can help brands engage customers in a more positive, personalized way. There’s money in getting this right: Bot‑based e‑commerce transactions are expected to reach $112 billion annually by 2023, according to Juniper Research.

Small wonder, then, why major retail brands are experimenting with bots that have the gift of gab. Whole Foods offers a mobile chatbot that acts like a personal chef, suggesting recipes and culinary inspiration via Facebook Messenger. It asks for your food preferences and dietary restrictions, and directs you to the right aisle for ingredients.

Sephora’s chatbot on Kik, an instant‑messenger service primarily used by teens, interacts like a stylish pal who suggests new cosmetics and offers makeup tutorials. Kik users can also upload photos to fashion brand H&M’s chatbot, which responds with personalized style recommendations. The bot can turn a shopping trip into an interactive social experience, letting teens share potential looks and vote on their friends’ outfits.

The bots don’t pretend to be human, but they can add lifelike grace notes that nudge customers toward transactions. When users tell Walmart’s new virtual personal shopper, Jetblack, “I need a birthday present for a three‑year‑old boy,” it eagerly responds, “My quest begins! How much would you like to spend?” Since Walmart launched the pilot program in May 2018, customers have reportedly accepted 79% of Jetblack’s product recommendations.


Adoption of AI in human resources is ramping up quickly. Chatbots are at the forefront of the trend, with the potential to reduce HR business costs by more than $8 billion by 2022, according to Juniper Research.

Enterprises are using automated chat tools to onboard new employees, provide training, and manage benefit requests. An employee can ask some HR bots, ‘Can I take this day off?,’ then have the request approved and automatically logged in the vacation calendar. “Chatbots are taking HR self‑service to the next level,” says Capterra’s Westfall.

As the chatbots become more conversationally adept, they can play a more important role in one of HR’s biggest challenges: recruiting and hiring.

For many job seekers, online applications are lengthy, frustrating and impersonal. Notifications on the status of an application are often vague or nonexistent; 43% of job applicants never hear back from prospective employers, according to Canadian career site Workopolis. Meanwhile, HR managers spend valuable hours screening resumes manually.

Conversational chatbots can guide candidates through the recruitment process, cutting down the time it takes to apply, and providing applicants with a more positive experience than receiving the standard “We received your resume” response.

The tools collect resumes and contact information and ask about candidates’ experience and skills. The bots rank candidates based on their qualifications, provide answers to frequently asked questions, and automatically schedule interviews for qualified applicants. “It’s a very pleasant, conversational way of applying for a job instead of, ‘I’ve got to fill out all these fields,’” Westfall says.

L’Oreal is integrating chatbots into its high‑volume recruitment process. The cosmetics giant, which collects more than two million online job applications a year to make 5,000 hires, uses a virtual recruiter from Mya Systems. Where an average human recruiter might spend 45 minutes on an applicant, Mya completes the process in five minutes. The chatbot even can ask such soft‑skills questions as, “Tell us about a project that you worked on that failed. What did you learn from that project?” Based on the answers, Mya scores candidates on their potential cultural fit at L’Oreal.

Still, chatbot technology has a ways to go before it can conduct a sophisticated interview. At this stage, companies should be wary of deploying it for late‑stage hiring.

“If you implement this too early and it provides a bad experience, that’s going to do irreparable harm,” Westfall warns. “Besides never applying to your company again, the candidate might tell their friends. ‘Hey, I had to talk to this broken robot to apply for a job. You definitely shouldn’t apply there.’”


In 2018, consumers logged an estimated 21 million chatbot interactions in the healthcare sector, according to a Juniper Research study. That number is expected to soar to more than 2.8 billion interactions by 2023, when healthcare will account for 10% of all enterprise chatbot conversations. Key to that growth is the rapid maturing of the technology’s conversational abilities, says Michael Larner, associate analyst for Juniper and the report’s lead researcher.

“Conversational AI is going to handle a lot more sophistication, possibly more than in other sectors,” Larner says.

Some companies are deploying virtual health assistants outside customer service. New tools offer diagnoses, recommend treatments, and provide virtual caregiving. They promise to relieve the burden on an overstretched healthcare system.

Babylon Health’s conversational chatbot uses speech recognition to ask about a user’s symptoms and, if necessary, refer the patient to a live physician. Florence and Sensely’s Molly are virtual nurses that monitor post‑hospitalization care, remind users to take medication, and track symptoms for complications.

Healthcare bots will only gain in acceptance as their conversational capabilities improve, Larner says. “You’ve got to make a natural conversation AI technology so that the user doesn’t even think about it—so it’s second nature to them,” he says.

Larner predicts that as health chatbots become human‑like, patients paradoxically will be more willing to confide in them than they would with live caregivers. AI chatbots will always respond sympathetically and nonjudgmentally, and the seemingly anonymous nature of the interaction will make patients feel more comfortable admitting that they haven’t taken their medication or have had another fall.

“You could argue that patients are more likely to admit these things with a chatbot than face‑to‑face with their doctor,” Larner says.