Chatbots, like their voice equivalents IVR, have long had a bad press. We’ve all had that sinking experience when you discover that a website’s chatbot only seems to understand about 5 key phrases, none of which have anything to do with your particular concern. And to add insult to injury you can’t find anyway to simply email the business to escalate your issue.
To be fair, businesses that implemented chatbots early on had to put up with limited functionality, and the requirement to hard code for every phrase they might be asked. And there was a heavy upfront cost to put the technology in place to service the interactions. Fortunately that is no longer the case. Natural Language Processing is available on a pay as you go basis with no infrastructure investment. And tools like Amazon’s Lex provide a framework for developing the interaction with your customers. Given this how can you move from simple chatbot interaction to proper Engagement Automation?
Make your chatbot seems more human
Make the way your customers interact with a chatbot seem more human. Machine learning, AI and natural language processing (NLP) mean that developing a chatbot is much simpler as you don’t need to code for every phrase some one may use. This ensures that chatbots can become more conversational. The underlying technology is constantly improving, so your systems automatically become better at understanding what is being said. And they provide feedback on phrases that are being used that don’t provide a answer. This allow you to constantly refine your model so that it will provide a more personalised and tailored customer experience.
Use sentiment analysis to steer the conversation
If you are speaking to someone directly, the way they are talking to you is as significant as the content of what they are saying. Say someone had missed a flight and wants to book another one. They may be pretty relaxed about it, as it was their fault, or they may be really angry because they were bumped off the flight by the airline. In the first instances handling the entire dialogue automatically may be no problem, but in for the second there are a number of actions you may want to take: automatically offer a discount/free trip because of the inconvenience. Or you may want to immediately hand the call off to an agent to have them handle the call.
Real-time sentiment analysis is now a standard feature of many platforms. With Amazon LEX you simply select it as an option when setting up the chatbot. All you then need to do is decide what you are going to do with the feedback provided.
Plan for the whole conversation
As suggested at the beginning of this blog one of the major frustrations of chatbots is only handling a limited number of options. So you need to really research the sort of conversations your customers want you to handle. Talk to your agents, analyse your emails and map out the topics you need to cover. With the sort of Agile techniques available nowadays for building out the conversation you should be able to rapidly create a model that you can test with your staff. And constantly monitor the responses, and continue to build out the topics (Intents) you can handle. Once you are happy to can handle a significant percentage of topics then you launch it to your customers. They will undoubtedly ask you questions that you didn’t expect, but these should be flagged up to you so you can continue to improve the responses.
Always have an easy way to escalate the conversation
Finally, make sure it is easy for customers to escalate the conversation if the chatbot doesn’t provide what they want. If you don’t want it to go to an agent then at least make sure they can create a generic email. Nothing is more frustrating than being sent to a webform with a drop-down that doesn’t allow you to create a topic you are concerned about. If it was a standard topic you wouldn’t need to escalate the conversation!
If you’d like to speak to us about Engagement Automation and chatbots then email us on email@example.com or call us on +44 20 7920 2000
You can read more about our Pay as You Go contact centre solutions here: JT Connect Contact Centres