Wolfgang von Kempelen was a Hungarian-born writer, physicist, architect and inventor who lived in the late 18th century. Although he brought several fascinating artefacts to life such as a waterpump, bridge, typewriter and fountains, he is most known for his fraudulent chess-playing automaton, The Turk. The Turk, which defeated many challengers including Napoleon Bonaparte and Benjamin Franklin, was a puppet of an Ottoman man sitting by the chess table with a fezz over the head. Believe it or not, The Turk is still accepted as one of the earliest Artificial Intelligence (AI) practices and all roads that lead to an effective contact center chatbot have to pass through a good understanding of the story of The Turk.
I can imagine how disappointed you are. Is a puppet with an Ottoman mustache, moving the pawn from one square to another by automatically following a predetermined pattern or responding to predetermined instructions really the ancestor of our nice and fancy chatbots? Well, yes and no! Before we get to a conclusion that early, let’s do what has to be done first and clarify what a chatbot is.
A chatbot is simply a bot (short for robot) that can chat automatically. It listens to you when you write “how are you?” and answers “good and you?”. The answer is predictable as it follows the instructions that its programmer programmed. Therefore, chatbots are far from being proactive. They cannot decide by themselves. This is how all Artificial Narrow Intelligence (ANI) softwares work. Once they are programmed to say “woohoo” each time Kim Kardashian posts a nude photo of herself on Instagram, they can “woohoo” and “woohoo” tirelessly no matter how often Kim breaks the internet (actually quite often). No matter how sensational the photo is, they stick to that defined expression. No “wow” or “meh”, always “woohoo”…
Let’s be honest, this is the type of personality that all contact center managers expect from their agents. Imagine an agent who is always polite to customers and does not change her pretty tone no matter how frustrating the customer or their problem is – an agent who can handle such a conversation 100 times a day.
– Welcome to Martin Airways! This is Mutlu speaking, how may I help you?
– I need to know if my flight is delayed.
– Your phone number is associated with Mr. Doe. Are you the owner of this number?
– Yes, I am.
– Ok! Your flight from Earth to Moon is at 1pm today and there is no planned delay. Is there anything else I may help you with?
– No, thank you!
– Thanks for choosing Martian Airways. Have a nice weekend!
Elon Musk is working on “travelling to the moon” part of the conversation. The rest of the conversation, speaking to relevant web services for more data about the customer, identity verification etc. is quite achievable with recent superior functionalities of computers. Their ability to handle multiple tasks at a speed that no human can achieve is what differentiates a computer from a human being. A computer can solve several triple integral problems while retrieving relevant information about your customers and their flights from your CRM solution and social media channels.
There Seemed To Be No Limitation For The Ability Of Machines
Although a pen was still the fastest tool to solve a triple integral problem at that time, Wolfgang von Kempelen lived in a noteworthy century. It was a time when machines found mass application areas for themselves. People were interacting with machines in all parts of their lives and they were amazed by what machines had done so far. There seemed to be no limitation for the ability of machines. They could even beat the great Bonaparte in a chess game. It was definitely the right time to work on a chess automaton. However, chess is a mind game and requires psychological manipulation to gain competitive advantage. It was not that easy to access all relevant data about the player and turn it into helpful insights for the automaton at that time. Thanks to connectivity and computer systems, it is a piece of cake for us now.
Wolfgang was a clever man. He used an extremely simple component to get over this problem. What he needed for his perfect automaton is what we need today for a perfect chatbot: a human!
The Turk was a mechanical illusion. It was operated by a human chess master hidden inside. The seemingly exposed innards of the cabinet did not extend all the way back. The hidden chess master slid around when the cabinet doors were opened and closed. With the help of cleverly arranged strings and wires, the chess master could easily operate all the actions.
Let’s reconsider the conversation between the airline agent and the customer.
– OMG, please help. I think I left my passport in the plane. I need to get in touch with someone before the plane leaves.
– Welcome to Martin Airways! This is Mutlu speaking, how may I help you?
– I told you already.
– Sorry, I couldn’t understand. May I ask you to say it again clearly?
– I said I lost my passport in the plane. I need to talk to someone in the plane.
– Ok! Your phone number is associated with Mr Doe. Are you the owner of this number?
– Yes… Oh God, I see that the plane is leaving *#&##&%
– Sorry, I couldn’t understand. Are you the owner of this number?
Boom! Our chatbot failed. Well, chatbots are computers and computers can only communicate well with other computers for now. Humans do not communicate in a robotic way. We have emotions that fluctuate depending on the context in realtime. Although computers are able to get information about that context too, they are far away from evaluating them in a human way.
Wit.ai is one of the most popular bot engines and it is used by more than 20k developers to build bots for different purposes. Before we blame poor user experience on our chatbots, it is worth taking a look at how we create them.
Bot engines use common “if this then this else this” statements to define scenarios. Input and the predetermined context are used as parameters to execute the output (answer). However, humans are semiotic animals. We process different signs, make meaning out of them and create meaningful communications. This is something beyond what can be achieved by “if this then this else this” statements.
There is no doubt that in the near future, empirical approaches will help chatbots get quite close to the way we communicate. Technology has proven to be limitless many times. We are now able to develop AIs that beat even the best chess players in the world. I also believe that agents will be replaced by chatbots one day, just like people in 18th century believed that a machine could beat Napoleon. Until bot technology is powerful enough, let’s do what Wolfgang did… Instead of building fancy but non-functional chatbots that pretend to be human, let`s use human-backed chatbots to help humans save time and effort.
Interactive Voice Response (IVR) uses DTMF and speech recognition to generate the input for the dialog. Given that, it is as easy as winking to create a logic that guides the flow of the conversation (Thanks to ECT`s powerful Visual CallFlow Builder). Once created, this logic can also be used by other channels such as text messaging, chatbots or mobile apps. Hence, all the functionality of IVR can easily be carried to different channels with no additional effort.
Additionally, chatbots can be used as Souffleurs to assist agents for the case of complex scenarios. During a conversation with a customer, agents can be suggested a relevant answer or helpful material by a chatbot that listens all conversations and learns from them.
Some Other Ideas To Leverage Chatbots And Create Meaningful Contact Center User Experiences
We the innovation team at ECT have some other ideas to leverage chatbots and create meaningful contact center user experiences. If you would like to listen to ours or share your ideas with us, please do not hesitate to give us a call. We would be more than happy to hear from you!