Chatbots are conversational interfaces, allowing people to communicate with machines. They represent a next-generation user interface (UI) and a more natural means of human-to-machine communication. Functionally, chatbots can accept written text from SMS, WhatsApp, Slack, Google, Facebook Messenger and many more, essentially, wherever someone or something is listening. The timing is excellent, as people are already using messaging apps with friends and family more than half an hour per day on average. Messaging is an accepted and comfortable form of communication. Additionally, businesses are now willing to leverage chat type interfaces, but there is a cost.
An interactive chat session with a customer service agent is costly, almost as much as the cost of a phone call, according to contactbabel. Enter chatbots, both a cost effective and technologically viable option. The technology required to support a more automated machine- or system-driven approach has reached maturity levels acceptable to mainstream business. Within the business world many customer support organizations have embraced messaging apps as a positive way to engage with customers. A modern chatbot strategy: conversational messaging driven by artificial intelligence (AI) with natural language processing (NLP) at the core, is the next evolutionary step for customer service and where organizations are choosing to invest early in the infrastructure required to support a chatbot strategy.
Designing and building a chatbot-enabled system of engagement, a next-generation customer services and support platform, is a journey that requires patience. To achieve success, business leaders must challenge the status quo—and work to build confidence and trust within an organization and with customers. The conversation is only one part of the problem: finding the answer is the second part. Bots can also be listening in the background and convert voice to text (think Siri or Alexa); but let’s solve one problem at a time and leave the voice-activated bots for a later discussion. We will limit this discussion to AI-supported text based chatbots.
The Future of Customer Service: Systems of Engagement
While there are many popular uses for chatbots, early business investments are focused on customer sales, service and support. Customer experience is the motivating factor. Customers want easy, effective experiences that elicit positive emotions when working with a business.
Customers are mobile. Messaging apps, driven by chatbots, will allow organizations to effectively engage with customers where they spend lots of time: on mobile devices. Customer service is among the better places to start the business conversation.
Understanding the relationship between a chatbot, artificial intelligence, big data and knowledge is foundational for this discussion. While that sounded oddly complex, all I really said was, ‘Listen, understand the question, figure out what the person means, find the answer to the question and let the person know the answer,’ all done by a computer.
Breaking it Down
The Conversation or Question
For a computer (system) to converse, it must have at the core an NLP engine and must also be able generate language: natural language generation (NLG). NLP and NLG use large data sets, aka big data, to determine the words, tone, context and type of question. Parts of this approach be outsourced to a service (Google, Facebook, Microsoft), but it is a use of both big data and AI.
Now that you understand the question, finding the answer is the next challenge. Some questions will be easier to answer than other questions. The answering process will require access to large structured data sets (Type 1) or unstructured data (Type 2).
Type 1, based upon structured information: ‘Where is my order?,’ a very popular retail customer service use case (it is a type of question). A chatbot receives an SMS reading, “Where is my order?” This process has a few steps in the back end (the fulfillment system) but should not be hard to imagine. ‘Look up the user, via phone number, look up recent order in order system, check status in shipping, send back response to chatbot, which makes the answer conversational.’
Type 2, based upon unstructured information or knowledge: ‘How do I?…’ or ‘What should I do if?…’ A simple example is spilling wine on fabric or your carpet. A chatbot receives an SMS reading, “I just spilled red wine on my new shirt.” First, the system may need to do everything from Type 1 above to figure out what the fabric might be, then converse with the user. The system may then need to search a knowledgebase or unstructured data, like video or other information stores. There is nothing particularly artificial here, but advanced search capabilities will be required.
There are many examples to choose from: to answer the question, access to lots of data and advanced computational power would both be required. This could be medical diagnosis or personal recommendations. Each requires the context of the question (location, language, emotion) combined with personal information, making more advanced requests of the system to ‘think’ and converse to determine a proper answer or course of action. There are nearly endless possibilities if the data (big data) is available.
Chatbots need not be just another headless UI
During the age of information (1990-2010), providing customers a way to search and find data via a web interface was good enough. Early search provided lists of results that needed to be sifted through in order figure out what to do next. Today, in the age of the customer, this is no longer acceptable. Customers want and deserve knowledge: information curated and modelled to be useful. Customers expect you to understand and respond to their request on the channel of their choosing, and for you to provide relevant and contextual information. Customers want answers; chatbots are here to help.
Chat usage rates increased significantly in the past five years. Messaging apps are becoming synonymous with chat. Both messaging apps and chat can and should be driven by chatbots. Chatbots are the more natural, intuitive, friction-free system of engagement, allowing for seamless transition between system interactions and agent. Extending your enterprise knowledge stores via advanced search is the key to success. In other words, understanding the question is only one part; being able to efficiently provide an answer is where the rubber meets the road.