Personal customer contact via intelligent chatbots: how it’s done. Part 2 of the series ‘Perspectives on Artificial Intelligence’.
Of all the current technological developments, artificial intelligence is both the most profound and the least understood. We are witnessing impressive new applications, but can hardly foresee their impact on people, organisations and society. In this series of blogs – Perspectives on Artificial Intelligence – we investigate not only the opportunities, but also the intended and unintended consequences.
Customer services using chatbots
Communication between individuals and businesses is increasingly taking place via chatbots. The simplest versions involve filling in a sentence or a range of key terms on a website, which then lead you to the list of products or frequently asked questions these cover. Only pre-programmed questions can be answered, and each user is given the same answers, as shown on the ING website below.
Another example is the Eneco Chatbot, allowing you to submit your meter readings via Facebook Messenger. This is done using a fixed script.
In such cases, should speech recognition (natural language processing) and artificial intelligence be added, the client will be given a very different experience. The chatbot will understand both the content and the context of the client’s questions and can react appropriately. The computer is able to ask follow-up questions, meaning a conversation can take place. With the most advanced systems, clients no longer know whether they are communicating with a human or computer.
The applications of such technology are countless. Take e-commerce, for example. Imagine reserving a restaurant table, making a new appointment with the hairdresser, or booking a trip. H&M use an application on Kik Messenger (similar to WhatsApp) which continues to ask questions and make suggestions until you have found yourself the perfect outfit.
And in 2016, Hilton – working together with IBM Watson – launched a robot named ‘Connie’. This robot informs hotel guests about local tourist attractions, recommends restaurants and gives information about hotel amenities.
Chatbots are becoming more human and more personal
Initially, chatbots can be trained using call centre conversations. After go-live, chatbots continue to learn from the answers given, as well as subsequent questions asked by the client with whom they are communicating. After all, thanks to machine learning, chatbots improve as they gain more experience. They can also learn to recognise and react to human emotions. An angry customer will be put through to a flesh and blood member of staff, while a happy one gets a different tone of response than an agitated client; to someone who indicates that the information given has not been fully understood, a more detailed explanation will be offered.
All communications between customers and businesses generate data which give insight into the wishes and motivations of the customer. This is valuable information, making it possible to segment clients more precisely, offer them appropriate deals and communicate with them at the correct level.
Self-service and chatbot acceptance is rapidly increasing
In 2016, research carried out by Aspect on 1000 Americans showed that 65% of consumers felt positive when able to have a question answered or a problem solved without needing to speak to a customer services staff member. In other words, what some businesses initially saw as a cost cutting operation turned out to have a positive impact on customer satisfaction.
With the condition that all technical functions operated to a high level, 49% of consumers admitted to preferring customer services contact solely by way of text messages. Younger generations especially have little need for telephone contact. However, some concerns remain; 60% of consumers say that businesses who use chatbots feel less human. Changes need time to become accepted; on the whole it is those customers who have not had experience with chatbots who say they prefer telephone contact with human personnel rather than with a computer.
And the staff?
Of course, the use of chatbots – or virtual customer assistants – leads to the displacement of customer services personnel. However, they will not be completely replaced any time soon. For the time being, chatbots are only used to deal with the simplest situations, while human staff members are allocated to exceptions and special cases. Chatbots listen in on these, learning what they should do for the next time.
Growing AI industry
Most of the new chatbots currently being launched are integrated into well-known social media apps, such as Facebook Messenger, Skype (Microsoft) and WeChat (the Chinese version of Facebook and WhatsApp). Via platforms such as BotMakers, you can contact businesses from various countries who can develop your chatbots at reasonable prices.
In the Netherlands, these developments are going forward at a clearly slower rate than in the United States and China. This is partially due to the Dutch language. CB Insights publishes an annual list of the 100 most promising AI start-ups. The businesses in the 2017 list collectively raised 3,8 billion dollars in funding. Few European players are included. Some of the more interesting businesses with respect to chatbots are Automat, Fido.ai, Kitt.ai, MindMeld, Narrative Science, Semantic Machines and Snips.
Step by step implementation of chatbots
Provided they are carefully and correctly implemented, chatbots can enrich the customer experience. Launching a chatbot is best done step by step. By starting with limited functionality, clients can slowly become accustomed to chatbot interaction. During this initial period, the chatbot can learn from the client’s questions and comments.
An important advantage of chatbots working via a pre-defined script is that it is easier to control the communications that chatbots have on behalf of your company. The self-learning chatbot Tay, launched by Microsoft in March of 2016, made racist comments on its first day of operation. You do not want your customer service chatbot communicating in such a way.
At present, chatbots are able to supplement, but not replace existing channels for customer services. It is important that clients can easily switch from the chatbot to a member of staff. This lowers the chance of clients getting lost and hanging up. Obviously, customer services personnel should be able to see the content of the chat, so that there is no unnecessary repetition and they can simply carry on from where the chatbot left off.
Earlier in this series: