Businesses are always searching for that killer idea, one that can keep customers happy while improving the bottom line. What if the answer is (quite literally) staring you in the face?
The expectations for chatbots have reached absurdly high levels. They cannot provide the companionship worthy of a starring role in the sequel to the Scarlett Johansson film Her, yet. But what they can do is revolutionise your customer support by answering customer questions 24/7 and improving the e-commerce purchase process.
Perhaps as importantly, the appetite for chatbots is there. In fact, two in three Millennials polled say they would use a bot to purchase products.
What is the recipe for the best chatbots?
The key ingredient in your chatbot is robust natural language processing (NLP) which allows the chatbot to understand human conversation – regardless of the slang and spelling used.
NLP is also important because it can recognize the intent behind a customer’s search. For example, if someone wanted to find out where they needed to check-in on an airline’s general FAQ website, they would be taken to a page displaying all the desks for every airline. However, a chatbot can use decision trees to find out exactly which airline they are flying on.
Decision trees are able to “botify” your FAQs to determine the precise answer. Observe the following example:
User: Where do I need to go to check-in?
Chatbot: I can find out for you. Which airline are you flying with?
C: Thanks! You can check-in at desks 26-32 in Terminal 3.
Decision trees provide simple questions which help narrow down the exact intent of your customer in order to provide the perfect answer. Even better, it is all done in a conversational format familiar to how the customer communicates in everyday life.
If NLP is the cake then what is the cherry on top?
Machine learning, combined with NLP, presents a powerful way to ensure your knowledge base is not only smart from day one but also will reach extraordinary levels of self-service as your data grows with more customer interactions.
For example, a customer might ask a ticket website, “Can I take my son to a rap concert?” It is possible that a knowledge base with only NLP will provide a FAQ result to the question: “What can I take to a rap concert?”
Linguistically, this is correct but not the perfect answer. As your data builds, machine learning will be able to learn from previous searches and supplant NLP in selecting the right question. Therefore, it might point the user to a different answer, such as “what are the age restrictions at a rap concert?”
This ensures that the learning curve for your knowledge base won’t plateau and will instead continue to grow higher than ever.
What happens if the chatbot cannot provide the correct answer?
This is where a hybrid chatbot can help you. In the rare occasions where a chatbot cannot locate the correct answer, it can transfer you to a human agent.
The bot will triage the support ticket which allows the agent to read through the conversation and understand exactly what the nature of the problem is which they can then solve. During the chat, the bot will still provide instant answers based on the question, allowing the human to select the correct one if they are relevant.
In addition, that support agent’s conversation can also be monitored by an expert agent who is able to step in at any time and provide further assistance. This ensures that on the rare occasion if a support agent needs additional help with a query, someone else can instantly support them. While chatbots with robust NLP provide 90% self-service rates, hybrid chatbots ensure that all customers walk away happy.
They might not be smart enough to become our best friends just yet but chatbots possess brains which can take customer relationships to a higher level than ever before.
Joe Lobo – Inbenta
Erfahren Sie mehr zum Thema im Webinar: