During SpeechTEK, which was held in late April in Washington, D.C., hundreds of customer care professionals gathered to wrestle with the shifting sand that is customer care, particularly the challenges and opportunities presented by automation in the form of virtual assistants or chatbots. The conference provided insights that included answers, objections, and more questions.
Companies spend massively on customer care and automation will reduce costs. In Tuesday’s keynote panel, Cognitive Code’s Chief Operating Officer (COO) Brian Garr said industry observers estimate customer care spending to be nearly $500 billion. Everest Group estimated in 2016 that call center spending was $300 to $320 billion annually. There is perhaps no other industry over the next few years in which artificial intelligence (AI) automation can impact a company’s bottom line as significantly as customer care. Virtual assistants can significantly reduce costs by replacing human agents.
In theory, consumers want virtual assistants. Intelligent virtual assistants tick numerous boxes in terms of demand-side customer care market drivers, including 24/7 availability, speed, and self-service. Virtual assistants could increase customer satisfaction and help companies overcome consumers’ widespread dread of engaging customer care. Samrat Baul, Senior Director for /7, said consumers want to use digital channels for customer service because their non-digital experiences are marked by too much time on hold and the desire for control, privacy, and an immediate start to service. But consumers want to know when they are dealing with a virtual assistant and expectations are different when dealing with virtual assistants/chatbots versus humans. According to Artificial Solutions’ Chief Marketing & Strategy Officer, Andy Peart, in his SpeechTEK presentation, when Artificial Solutions’ client Swisscom identified its chatbot as a machine, customer satisfaction increased.
Customer care requires expensive humans for a very good reason. Customer care issues are very complex because humans must explain their issues in their own words (spoken or typed), which then must be interpreted correctly to be solved. Natural language processing (NLP), an umbrella term that describes a family of natural language (NL) technology, which includes NL understanding, NL generation, etc., has improved drastically over the last 3 years, igniting renewed hope that virtual assistants can better understand customers.
Natural language processing combined with machine or deep learning helps develop meaning. Intent and context are very difficult for computers to discern. Computers do not easily understand sarcasm and other nuanced sentiment. Companies like Inbenta combine NL with machine learning (ML) and deep learning (DL). “Deep Learning improves our system,” said Jordi Torras, Chief Executive Officer (CEO) of Inbenta. “One side is pure logic; it knows the world. The second side is pure machine learning. The second side learns logic from the first side and now has experience based on history.” Inbenta’s system performs gap analysis on user questions, learning how customers ask questions, enabling the system to retrieve better answers more quickly.
Natural language processing systems require real data to train on, which is a problem. There are no beta tests for NLP systems because the system cannot anticipate how queries will be asked or how to classify those queries according to the right information sources, so systems go live and must learn on the fly. Some systems require several days before they are satisfactorily accurate, giving pause to many companies looking to deploy virtual assistants. But training times are getting better. Tractica spoke with CEO Dr. Catherine Havasi from Luminoso, who said the company successfully trains virtual assistants much more quickly these days. “Companies don’t know what kind of conversations they want the virtual assistant to have,” said Havasi, “So you need to understand what end customers want. Our system can do that in about 15 minutes.”
Breakthrough implementations are here. Cyprus-based customer care specialist Omilia impressed the SpeechTEK audience with a live demonstration of its cross-channel intelligent interactive voice response (IVR) solution, a totally unstructured user interface (UI) with no menus. Dimitris Vassos, CEO of Omilia interacted with the system using a wide range of tricky contextual questions, like “Do I have an outstanding balance on my card?” Followed by “How much is it?” And the system knew the second question referred to the first one. The system went live for Royal Bank of Canada credit card services in October 2016 and can be accessed via phone and Facebook Messenger.
Automate when and where it makes sense. Chatbots will not be taking over the customer care world anytime soon. Best practices are forming and experts understand that customer care for most companies will always require a human touch. Hybrid systems where seamless handoffs between humans and virtual agents are implemented will continue to grow.