How conversational AI concepts enrich customer service interactions
By Fredrik Tunvall | 4 minute read | September 5, 2018
- Virtual assistants are more advanced than they used to be. They are no longer linear but instead use machine learning and intents to communicate more naturally with your end users.
- There are five core conversational AI concepts that will help you to build a virtual assistant for your business: User utterance, Intent, Entity, Dialog and Dialog Nodes.
- AI assistants like Watson can understand context, and enhance your customer service.
When your customers seek answers to their FAQs, they aren’t working from a script. Clients will typically use the same language with an automated customer support system than they would if they were talking to a live agent. This means that they could use expressions, slang terms, or verbiage which differs from that which you used to train your chatbot.
Conversational AI allows for nuanced interactions which are more human-like than typical chatbots. They don’t just follow straightforward “if-then” flows; they incorporate natural language understanding to make human to machine conversations more like human-to-human ones. Building an AI assistant to handle common customer support questions will allow you to communicate more effectively with your end users. But, before you start building your solution, it is important to understand the core conversational AI concepts in order to start development. Let’s break down the concepts in the below customer service scenario:
It’s 3 a.m. on a Tuesday. Samantha logs into her cell phone service provider’s website, looking to switch to an unlimited data plan. She has been loyal to her carrier for as long as she can remember, but after seeing a competitive offer at a more affordable monthly rate, she wants to investigate her options. Shortly after Samantha’s account page loads up, she sees a slightly more expensive unlimited plan. A customer service chat window pops up.
“Hello, Samantha. This is the Account Services bot. How may I assist you tonight?” reads the message.
“Hey,” Samantha types in the interaction window. “I just saw an ad from a competitor offering uncapped data plans for cheaper than what I’m paying per month. Any chance you guys would match that?”
A reply message quickly surfaces on the screen. “Sorry, Samantha. We do not offer uncapped data plans. I apologize for any inconvenience this may cause. Please try us again after 9 am Pacific Standard Time, and a live agent will be more than happy to assist you further. Was there anything else I can assist you with?”
“No thanks,” Samantha replied. She shuts her browser and decides that switching to the other carrier is her best option.
Conversational AI concepts which drive innovation
User utterance – In this scenario, the question Samantha’s asked the virtual assistant was the user utterance.
Intent – When Samantha interacted with the virtual assistant, she wanted to find an unlimited data plan at an affordable price. That was her intent. Intents are purposes or goals expressed in a customer’s input.
Entity – Samantha was calling about her data plan, which we call the “entity” of the conversation dialog. An entity represents a term or object that provides context for an intent.
Dialog – How an AI assistant processes the conversation, including the user intent. The dialog of the discussion in this scenario was limited to the virtual agent’s inability to deviate from its linear programming. A dialog can be through online chats, SMS interactions or voice conversations on the phone.
Dialog nodes – At a minimum, a condition (or question) and a response. A dialog may include multiple nodes, such as when an AI assistant is sophisticated enough to ask clarifying questions to understand a user utterance.
The Evolution of Conversational AI
In the early days of customer service virtual assistants, a user would ask a question, and the query would be submitted, relative the linear rules-based instruction set the chatbot was trained on. It could only serve the customer based on a script in its coding. It couldn’t apply context, like whether a website browser was a long-time customer, or whether a user was from a region with cultural dialects or speech patterns. Conversations on one channel, such as an online chat would only be captured within the system that managed that mode of interaction.
AI platforms have evolved, and can now learn from, and apply context to multi-modal conversations. Companies which augment their client-facing operations with AI assistants can be integrated with core systems like CRM, to identify high-value customers who should receive extra consideration, such as loyal customers that have been long-term subscribers.
Modern conversational AI platforms allow for contextual bias, can understand a broader understanding of human language inputs and can even assist customers to complete transactions based on their advanced ability to guide a customer through a conversation. If the Account Services bot offered Samantha an unlimited data plan without switching providers, she likely would be happy to stay loyal, even with a difference in monthly cost.
Not all AI platforms are equipped with natural language understanding ,or can understand the nuances of casual conversation nuances. Conversational AI will encourage your users to get answers to their common FAQs, leaving the more complex queries for your live agents.
To learn how to get started building a conversational AI solution for your business, register for the IBM Masterclass: Core Conversational AI Concepts. This is a 6-part video series that will teach you everything from why you need conversational AI to advanced conversational AI concepts.
Watch Episode 4 of the Masterclass.