Natural Language Works with More Than IVRs
The first attempts at computer-based natural language understanding (NLU) predate the debut of the personal computer, and just like the PC, NLU technology has advanced tremendously in the decades since.
With data available to train systems and computer power growing exponentially over the past several years, NLU, combined with natural language processing (NLP), can understand and reply to what a person is saying or typing in milliseconds, empowering interactive voice response (IVR) systems and chatbots to efficiently handle an increasing number of customer interactions.
Recently, it’s the increased use of the natural language technology across an array of customer interactions that has evolved much more than the technology itself, according to experts.
NLU is being used to provide easier ways for customers to interact with companies via a variety of channels, says Sebastian Reeve, director of market development for intelligent engagement at Nuance Communications, which was recently purchased by Microsoft in an $19.7 billion deal.
While natural language technology is most often associated with speech and IVRs, consumers are increasingly using texting platforms, so NLU is increasingly being applied to those channels as well. It is also being applied more frequently to assisted service, advising contact center agents on callers’ issues to make the outcomes that they deliver more consistent.
And while the technology was on the rise prior to the pandemic, COVID-19 caused an explosion in self-service, and NLU and NLP proved critical in helping companies handle the surge, according to Daniel Ziv, vice president of speech and text analytics and global product strategy at Verint. “You need solutions that can understand intent and drive automated responses. To figure out what those questions are, you need solutions like speech and text analytics that have NLU embedded in them to mine the incoming conversations in unstructured voice or text.”
Amid the pandemic, NLU was also a key enabler for companies to make better chatbots to handle the influx of customer interactions, agreed Steve Levine, chief marketing officer at Cortical.io, a provider of natural language technologies for mining text messages.
Chatbots, Levine explains, were historically rules- or tree-based, and if the customer didn’t follow a specific path, the chatbot would fail, either leaving a customer hanging or pushing the customer to a live agent. “Adding natural language understanding capabilities to these chatbots reduces abandonment and enables customers to achieve the goals they want.”
In addition to reducing abandonment, NLU also enables chatbots to correctly handle a larger number of interactions, meaning fewer are handed off to live agents, Levine adds.
NLU: A TRIAGE MECHANISM FOR CUSTOMER INTERACTIONS
Another advantage that NLU provides for customer interactions is that it can distinguish simple information requests from urgent requests, properly prioritizing and routing the information. For example, a customer or prospect requesting an insurance quote can be handled automatically or forwarded to a salesperson, depending on where he is in the sales cycle. If the call or text message, on the other hand, is about a car accident or other emergency, NLU could escalate the communication to a live agent.
In the past year, companies have focused on using that kind of data to retune and recalibrate and add more digital self-service options, Ziv says.
One large telecom company, for example, applied NLU to its self-service bill payment feature, which led to a 10 percent reduction in call volume, according to Ziv.
“NLU provides a richer context than traditional mechanisms,” Reeve adds. If, for example, a customer uses an IVR with five menu options, the company has no more context than which option was chosen. But “natural language tells you quite a lot of other things,” he states. An IVR might give the option of pressing 3 to get information on overseas use of a credit card, but with natural language, a consumer could say something like “I will be traveling to Italy and want to know if my credit card will work there.”
The deeper context offered by NLU enables the credit card issuer to offer more specific and personalized travel insurance as well as other cross-sell opportunities via chatbot or live agent.
“I can highly tailor personalized experiences,” Reeve says. “It’s a way to generate revenue. We’re able to drive other algorithms, like prediction and personalization, which thrive on contextual relevance. It starts to unlock a lot more use cases and fit for both the brand and the customer.”
Fashion retail giant H&M now enables consumers to engage with its virtual assistant and live chat agents directly from services like Google Maps or Google Search. Through an integration of Nuance’s virtual assistant and live chat, H&M can leverage prior investments in Nuance’s AI-powered Intelligent Engagement platform to give customers more choice and flexibility when shopping online.
Nuance’s latest integration with Google’s Business Messages further allows H&M to take advantage of new messaging channels while leveraging the same AI engine at the core of its successful virtual assistant and live chat deployment. The result is more seamless cross-channel customer experiences, reduced contact center call volume, and greater digital interaction.
Such use of NLU also improves contact center efficiency, according to Nuance. Every year contact centers receive billions of calls originating from Google applications. An NLU messaging option that integrates with the existing virtual assistant and live chat deployments allows customers to select their preferred channels, which can significantly lower an organization’s overall call volume.
NLU enables companies to understand and analyze the topics of conversations and the emotions expressed during them, says Julie Miller, vice president of product marketing at Clarabridge, an interaction analytics provider. Companies can then use that information to change business processes, improve products, and provide better customer service.
“The biggest shift that I’ve seen over the past three to five years is the growing dependence on interaction analytics and a growing dependence between customer experience teams and the contact center that has all the data,” Miller says.
Contact centers have long had the embedded information from recorded calls but used it primarily for training. Now they are sharing the NLU analysis with CX teams to improve customer interactions, Miller adds. “CX teams are looking for new sources of data because customers are suffering from survey fatigue. At the same time, contact centers are waking up to the volume of data that they collect.”
Ziv also points out that the pandemic led to an increase in highly emotional calls from distressed customers worried about the virus, uncertain working conditions, family issues, home schooling, and elderly relatives they couldn’t visit. So Verint concentrated on updating its human-to-human conversation with real-time guidance and launched several initiatives, including a new real-time agent assist solution that is much more contextually aware and accurate. It features linguistic, acoustic, and desktop triggers that analyze what’s being said, how it’s being said, the sentiment involved, periods of silence, interruptions, and more.
HEALTHCARE USE FOR NLU INCREASES
The use of NLU increased across many industries last year, but perhaps none was as profound as the particularly large growth in the healthcare sector as providers sharpened their focus on value-based care and patient experience. Government programs such as Medicaid and Medicare pay very close attention to patient reviews, so the natural language technologies become increasingly important for understanding patient communications and providing superior experiences.
“Healthcare adopted it aggressively,” says Dakshi Agrawal, an IBM fellow and chief architect for AI. “CVS used our NLP product—which includes NLU—to help handle their call volume as it rolled out its vaccination program.” Others offering vaccines, like Kroger, also used the technology, Agrawal says.
Retailers also stepped up their use of the technology during the COVID-19 pandemic, particularly as they rolled out order-online/pick-up-in store initiatives, according to Miller. Similarly, NLU enabled retailers to aggregate data from conversations on social media and other channels to help identify stores that had shortages of toilet paper and other essentials so that deliveries could be routed accordingly.
Agrawal says natural language technology provides the best return on investment in industries like healthcare, telecommunications, and financial services, which have high customer interaction volume and a large percentage of interactions seeking the same types of basic information, like account balances.
“This support becomes invaluable when you have millions of customers seeking information,” Agrawal states.
Ziv also notes that NLU is being used to help detect and understand differences in communications unique to different channels. All caps in a text message could mean anger or urgency, or it might just be that someone accidently pushed the caps lock button. Natural language processing and understanding can look at the text and the context within the message to determine the correct meaning. Similarly, pauses or shouting in a voice message can have different meanings. One customer might be yelling because he’s angry; another customer might just be hard of hearing.
NLUS SERVES AS A FRAUD DETECTION AID
Besides aiding with self-service and agent-assisted customer interactions, some companies are turning to natural language technology to help detect fraudulent customer interactions, according to Sean Drummy, senior director of product management at Vee24, a provider of AI-enabled customer engagement systems.
This has been particularly true of retailers that ship larger items, such as furniture.
“Natural language understanding helps you detect an outlier for really any purpose,” Drummy explains. “You could similarly utilize natural language understanding to identify markers of a conversation where this individual may be trying to do something fraudulent.”
Conversations with a fraudster tend to have certain tell-tale signs, which NLU can recognize and flag, Drummy adds. If, for example, certain payment apps are commonly used by fraudsters trying to convince someone to send them money. NLU can flag when those apps are mentioned in an interaction.
“Generally, the strategies employed by these people all have the same tact,” Drummy explains further. “So anytime you’ve got a pattern that you can establish using NLU, you’re cooking with gas with regards to detecting fraud.”
LOOKING AHEAD FOR NLU IN CUSTOMER SERVICE
Half of all businesses today have already adopted applications powered by NLU or are planning to adopt them in the next 12 months, according to Agrawal. “In terms of growth of NLU for customer service, this is going to just accelerate digital transformation.”
Additionally, the IBM Global AI Adoption Index 2021 noted that “increasing the level of adoption for NLU/NLP technology in the future will hinge upon how businesses make use of new tools that automate many of the common barriers to entry—for instance, lowering the requisite skill set for training and deploying language models.”
The IBM report also noted the top five barriers to entry for natural language technology:
- the high cost (29 percent);
- the amount of training required (26 percent);
- difficulty keeping the technology up to date (24 percent);
- complexity of use (22 percent); and
- a lack of the requisite skill set within an organization (22 percent).
Cost was cited as the greatest barrier to adoption in the United States, Latin America, and Europe; however, training requirements are a greater barrier in India, while complexities and lack of customization are barriers in China, according to the report.
But Nuance’s Reeve expects NLU to further evolve as companies continue to train it with conversations gathered across channels.
“The way that we speak, the way that we type, the way we talk to Alexa versus a phone-based IVR are all different,” he explains. “We’re going to start seeing organizations wanting to have a single team developing conversational experiences across all of the touchpoints that they have with their customers.”
And once that happens, there’s no telling where the technology will go.
Phillip Britt is a freelance writer based in the Chicago area. He can be reached at email@example.com.