Talk about bots is everywhere it seems. But as the old technology adage goes, “You shouldn’t use technology for technology’s sake”, otherwise, you give the technology a bad name. While that is typically how technology adoption starts —as someone has to go first, when it comes to customer-facing applications that ostensibly replace or augment people, a measured approach is required.
We are quite a ways from a connected world in which AI will drive our daily lives through connected homes, autonomous driving cars, virtual agents, and intelligent personal assistants, in particular, those that talk and pass off information to each other, as the norm rather than lab projects. However, the table has been set for this to happen. In customer care, the intelligent personal assistant precedent was set years ago when we first added speech recognition to Interactive Voice Response (IVR) systems. Yet the case for mass proliferation is still far off and being driven one application at a time.
In order to reach the state where we move from lab to status quo three things need to happen:
- A significant installed base of VAs and Bots have to prove their worth, with measurable results in increased Customer Satisfaction (CSAT), a reduction in customer effort, and a solid return on investment (ROI).
- Industry wide best practices in design and implementation must be developed and continuously refined (not to mention adhered to).
- Bots and VAs need to emerge out of an overall plan for customer engagement that focuses on improving the Customer Experience (CX), not as one-off applications or to be cute.
So let’s break down the bots as they apply to bullet one. Starting with the core of what this author likes to talk about, customer contact, Bots, and their namesakes—, virtual assistants, intelligent personal assistants, and other naming variants—are making inroads into the digital workforce and producing measurable ROI and an improved CX.
AI-Infused Classic Virtual Assistants
The word classic is used here, as the initial concept behind speech-driven virtual assistants was to offload agents in a contact center. Out of this came numerous companies that have been around for a decade or more, initially or solely focused on virtual assistants. In addition, many of the legacy suppliers of contact center technology have either partnered with virtual assistant companies or have produced their own, such as Avaya’s AVA, Genesys’ Kate, Nuance’s Nina, or Aspect’s Mila. Here are a few virtual assistant examples which are producing measurable results:
Much has been made about IBM’s Watson, and in the case of Autodesk, Watson is doing exactly as advertised in the context of agent replacement and assistance. As reported in Venture Beat, Autodesk went all in back in June of 2016, by piloting the IBM Watson Conversational Service. Now branded AVA (Autodesk Virtual Agent), this virtual workforce is capable of handling 80% of customer support contacts for both voice and web, allowing customers to interact through chat to get their issues resolved.
AVA is powered by natural language processing (NLP) and deep learning, and was built through analysis of a base of over 14 million sentences, gleaned from various customer interactions repositories including chat logs, forums, and use cases. Internal subject matter experts also are enhancing the knowledge database with their own information, as well as monitoring and tuning Ava’s algorithms and vocabularies to improve the solution over time.
According to Autodesk, the return on investment (ROI) has been outstanding, with a 99% increase in speed of inquiry handling, from an initial average of a day and half to resolve to only five to ten minutes. If AVA can’t resolve a situation by itself, it collects enough information to create a case and forwards the ticket to a human agent. Autodesk also reports a rise in customer satisfaction as a result.
NextIT and Dell
Next IT is a conversational AI company that initially had its roots in virtual assistant technology, but over its tenure has developed language understanding components and customized conversational AI that fits into every type of conversational solution on the market. The company views bots, virtual assistants, and other solutions as simply endpoints where they can insert intelligence , and supports solutions within customer contact as well as other areas of the workforce. This intelligence is powered by a data library that is curated from the real-world and validated by experts to include over 750K unique terms, more than two million symbolic patterns of terms, more than 90K business intents, and over 165K unique actions.
One of the compelling features of the Next IT product stack is its Author AI, which is a Content Management System that allows a business to apply the same language model across different conversational solutions, enabling a developer to develop once and deploy everywhere, and the business user to modify the responses, no matter the form factor. This enables a company to launch Web and mobile apps, across the spectrum of VA/bot types, from Twitter, Facebook Messenger, live chat, SMS, or in-home assistants such as Amazon Echo.
Putting this into real world usage, Next IT worked with Dell.com to create a chat bot, Ava, to increase customer satisfaction (CSAT) and reduce costs by offloading live chat resources. AVA is a virtual sales agent that can provide navigational assistance on the Dell.com Web site, make product recommendations and find deals, or access information such as technical specifications for customers. She has understanding of all the catalog, product and pricing information on the site. Ava is a supplement to live chat, and is available across platforms including the Web, mobile app and tablet.
Ava, who has been working at Dell for about six months, has already proven herself an energetic worker. In just three months, chat instances for both live chat and the IVA contact rate increased by 61%. At the same time, live chat costs decreased by 27%. Ava was able to assist customers satisfactorily and close conversations at a greater rate than live agents alone, while also reliably driving revenue with product recommendations.
Ava is not only able to effectively assist customers in navigating the ever-changing site and instantly support questions, but also accurately recognizes customer intent so well that it is able to compare ever-changing product selection by understanding language about new products and technology before customers do, in order to advise on purchasing decisions.
Inbenta and Ticketmaster
Ticketmaster is one of the largest e-commerce companies in the world, and one of the most well-known global brands in the area of entertainment and ticketing. With annual revenues of $8B, the company supports 6500 ticket locations, and 7000 agents in 19 worldwide sales and customer care centers.
Ticketmaster set out to tackle the twin challenges of improving customer service, while reducing costs when it brought Inbenta on board to enhance its self-service option. The goal was to deploy a natural-language-based, intelligent assistant chat option that could assist customers in more quickly finding answers before they completed transactions, lessening the impact on its customer support centers. It also wanted to dynamically improve its FAQs with the results to further improve its knowledge base.
One of the attributes that drew Ticketmaster to the Inbenta solution is that it reduces the time it takes to assist customers by searching on meaning not keywords. This allows it to populate the chat screen with possible search results as the customer is typing, upping the chance that they will more quickly get answers, and lessening the chance of calling or sending an email to customer support.
While providing service to customers in 19 countries and in 14 languages was the ultimate goal, Ticketmaster first started with a four-month trial in the United States to compare year over year self-service rates. The trial resulted in a 42% improvement in self-service, increasing its self-service rate to 96%, with an estimated savings of $550K a year. One of the notable statistics from the trial included a 22.8% reduction in emails being deflected from its contact centers. Overall, the solution also reduced the agent fatigue that comes from repeatedly answering the same questions again and again.
InBenta currently supports 25 languages. Ticketmaster now has the solution deployed in more than 20 of its websites, in 30 countries and multiple languages. While the self-service application gets better over time, Inbenta also developed additional contact forms for agents to gather more information for support agents. You can check out the assistant on the Help link on Ticketmaster’s site.
The Last Word
As a final word, let’s circle back to bullet number two. Here are some of the core design questions you should ask yourself before launching an AI-infused application:
- What is it that your customer is trying to do, or what is it that you want to enable them to do, that they can’t right now without live assistance. Sounds simple, but this is the most basic question you should ask. What is it that can be solved by a virtual assistant over a live one?
- Is it appropriately placed in the customer journey? When chat bots first were introduced, they just popped up everywhere, and could be annoying, or simply ended up as a lost opportunity for customer engagement at the point of need, as customers clicked on “x” to say no to chat, so they could go about their business.
- When is the human touch required? AI lacks many human qualities, such as empathy, judgment and creativity. The industry has spent years fine-tuning when and how a supervisor should step in to assist a live agent when an interaction is going south. Don’t expect bots to be better.
- Are there areas of the business where a bot could make an impact that aren’t customer facing? The answer is certainly. Robotic Process Automation (RPA) has been around for years, and is making inroads to customer contact. RPA can be used to automate countless areas of the back office that impact customer care, from pre-processing mortgage loans, to offloading accounting functions, shipping and logistics, and other areas.
But companies should be aware that RPA comes in many flavors, the majority of which do not incorporate machine learning and don’t learn over time, but are instead task-driven. This is all the more reason to build out an AI strategy (bullet 3), if you intend for the Bot in the back office to increasingly improve.
