Virtual digital assistants are starting to pop up on the websites of online retailers and service providers. These digital agents, trained in the gift of gab and often mimicking a live human agent, engage with web visitors to answer simple questions or point to relevant web content. Common uses of virtual conversational agents include providing web visitors with information around fairly static business practices, such as an online retailer’s return policy or information about shipping.
How effective is the current generation of virtual agents? This is a question I’ll be exploring on this blog. To start the ball rolling, I think we can safely say that conversational agent technologies has made marked advancements over the past decade, but it still has a long way to go to completely replace live agents offering deep levels of customer support and interaction.
The capabilities of a virtual agent are largely determined by their underlying technology. The simplest digital agents, often referred to as chatbots, are little more than pattern matching programs that look at user input and then pair that input up with a predefined matching response stored in a large knowledge base. Chatbots are notoriously inflexible, since they are only able to respond to input that matches one of the patterns in their knowledge base.
More versatile virtual agents take advantage of natural language processing engines that allow them to recognize a variety of inputs. As with chatbots, these agents still require a knowledge base of appropriate responses that they can use to reply to input, once they have construed its meaning.
Currently the sweet spot for conversational digital agents seems to be in use cases that mirror Frequently Asked Question scenarios. If the question being asked by the web visitor is a common question, and if the proper response is static and easily stored in a knowledge base, then virtual agents have high rates of success in answering the visitor’s question. As soon as the questioners deviates from the known lexicon of commonly asked question, however, the chatbot or virtual agent is obviously out of its depth.
In future posts, we’ll take a look at some chatbots and put them through their conversational paces.