Chatbots: Beyond the hype, now what?
The time when we start with chatbots ‘because it can’ is over. Instead, we look at data-driven chatbots that will really add something to our business processes. Chatbots will become a stable factor in our developments and with this we will confidently enter the fourth industrial revolution with artificial intelligence, robotization, Internet of Things, Cloud Computing and 3D printing. Or not…..? How has this hype developed in recent years? And how should we continue from here? In this blog, we will take you into account in all developments.
How did the chatbot hype start?
Where we used to see many opportunities and possibilities in the field of chatbots and Artificial Intelligence, Gartner now expects that the hype has passed its peak and will decrease. Now that we’re almost leaving these ‘Peak of Inflated Expectations’, we are entering the third phase of the cycle. This is the phase in which we see that our high expectations need to be nuanced somewhat (Trough of Desillusionment) and we need to draw learnings from practice, so that we eventually get ready to make the technology productive and yield a return.
Where hypes often show an enormous growth in a short period of time, we actually know chatbots for much longer…
Chat bots are not new
We have known chatbots for years, but it took a while before we could make them ‘smart’. It all started in 1966 with Eliza, the chatbot that was the first proof of superficial contact between man and machine. Later, in the age of MSN, they were also indispensable. Think for example of the chatbots who entered into conversations with you about all kinds of things. And so there are still many bots that have had an impact on the development of chatbot technology as we know it today.
These bots were in no way ‘smart’. After all, they responded with standard answers to messages that were recognised based keywords. The development of technology has now gained momentum in recent years, our chatbots are becoming smarter and virtual assistants, such as Google Assistant and Alexa, are becoming increasingly accessible to consumers.
From overwhelming chatbot success to latent disappointment
Thanks to the technological possibilities of the past few years, chatbots grew like mushrooms. Anyone could develop a chatbot. We saw many applications passing by via Facebook Messenger, when Facebook opened up the platform to developers of chatbots.
Were we cheering too early? Because let’s be honest, the applications were not successful and were mainly experienced as irritating by users. These bots are programmed to recognize certain keywords, but do not ‘understand’ the customer. Sometimes the customer ends up in an infinite loop that does not answer the question asked. Research even shows that 70% of all bots on Facebook are unable to successfully perform a simple task.
Not because it is possible, but because it must be
But is it now time to throw in the towel and give up the promised success of chatbots? Absolutely not. The hype cycle has passed its peak, but now we can learn from the successes and mistakes of the past. We are building a stable future for chatbots within our processes.
Organisations no longer look at the possibilities of technology ‘because it can’, but rather look at the added value of data in this process. Data can predict what the success of a chatbot will be for certain tasks. What it delivers in terms of efficiency and cost savings.
Please note! A chatbot is not suitable for all organizations. Therefore, make sure you perform a chatbot scan, which can tell you what the expected success of a chatbot is. Analyze response times and determine where you can win on first response, lead time and first time fix. Or look at peak moments during failures in the past year. Does the scan show that a chatbot is not suitable for that specific task? Then adjust the process and see where a chatbot can be of added value or terminate the process in time.
Chatbots in practice: smart through Artificial Intelligence and Natural Language Processing
But how does a chatbot ever really engage in conversation with people? The question that arises in many heads. The answer lies in Artificial Intelligence and Natural Language Processing. A bot ‘learns’ the language of the customer, based on conversations from the past or input from people. In addition, he learns to recognize the intention of the customer and which data is included in the customer’s message. Then, based on business ruling, the bot examines what his next question should be. He requests missing data from the customer or escalates to an employee if necessary.
Chatbots for online customer service
Will chatbots ever completely replace humans within service? No! The secret of a successful application lies in the perfect mix of technology and people. Always keep in mind that a chatbot should contribute to the customer experience. Prevent irritation and ensure that a chatbot seamlessly connects with your customer, but also make sure that your chatbot does not get in the customer’s way and passes the question on to the employee if that benefits the conversation.
Besides chatbots drastically reducing the response time and lead time of conversations, adding technology also increases customer satisfaction. Customers with a simple question expect a quick response. With a chatbot, this reaction can be there within seconds. It also increases employee satisfaction, because the work of a customer service employee becomes more challenging. More complex questions remain and employees can be used differently and better to take an extra step for the customer.
The ROI of chat bots at HEMA and bol.com
Figures from HEMA and bol.com, for example, show that chatbot technology has proven itself in the field of online customer service:
– HEMA made the choice to have messages about an online order or ‘more HEMA’ caught by a chatbot. Via Facebook Messenger, this concerned 29.4% of all messages. In 17.2% of all these messages, no order number or card number was sent by the customer. The chatbot took action to request the additional information from the customer, so that employees could immediately start working. This reduced the processing time for such questions. HEMA saw a cost reduction of 17%.
– Bol.com works with a chatbot that works seamlessly with other chatbots and with people. This bot is currently working on Facebook Messenger, but it can also be used on other channels. He has the task of work preparer, where he requests specific information from customers, such as an order number or an e-mail address. The Service Expert can then immediately get to work on this. In addition, when a customer asks a question that is suitable for chatbot Billie, he can refer him to Billie. In this way, the chatbot can pass on customer demand to both chatbot Billie and one of the Service Experts. At bol.com 36% of the costs have been reduced and employees have 20% more customer interactions per hour. The use of the chatbot has freed up time for specific customer cases.
To bot or not to bot; are you ready for the future?
When developing chatbots, the customer experience should always be the important thing. That is why it is important to ensure a seamless collaboration between people and technology. A bot works either behind the scenes to ease the work of an employee, or in front to directly help a customer. In the latter case it is important to think carefully about business ruling and moments when the chatbot escalates to an employee, in order to avoid frustration and repetition at a customer.
Are chatbots suitable to be used for customer service? Absolutely. But keep in mind that customer contact will never be fully automated, because there are always customer questions that need an empathic response. So make sure you have the ideal mix between people and technology!