Not everyone was initially enthusiastic about chatbot Iris. “(Online) service is simply not in the DNA of every ANWB employee,” says Margot van Leeuwen, Customer Experience employee at ANWB. Colleague Carolina van den Hoven adds: “At the end of the day, we are a commercial organisation. So, how do you find a good balance between sales and service?”
Iris was tasked with one main goal: to reduce the volume of contact in the Member Service Centre by answering common questions. Over the years, this focus has shifted from contact reduction to increasing online customer satisfaction.
As ANWB now has a large proportion of customers reaching out to Iris for help, ANWB is continuously ensuring that Iris provides a high-quality service.
"It took some time and effort to get all of our colleagues enthusiastic, but we see that they are now more and more involved," says Margot. By believing in the power of Iris and continuously improving her throughout the years, there is no longer any doubt about the added value of Iris when it comes to service. Summer 2019 proves that.
Every organisation has its peak periods. For ANWB, their peak period is the holiday preparation period, when customers are asking questions about roadside assistance abroad, travel insurance, toll badges, environmental stickers, credit cards, and much more.
To increase customer satisfaction, ANWB wanted to ensure that Iris was recognising most of these questions during the peak period and could answer them correctly. Therefore, the Customer Experience team decided to invest some time in thoroughly analysing and optimising all of their content. During this optimisation phase, ANWB followed three principles.
We often tend to make assumptions based on our own ideas and experiences. Margot: “You have to let the idea go that you know what customers mean by their questions. You just have to delve into it: what does the customer really want to know?”
ANWB investigates this through internal workshops with employees from different departments. During these workshops, customer signals are examined to find out the essence of questions. By looking at customer questions through all the different channels, such as Google Analytics, the Member Service Centre, and of course chatbot Iris, you gain a complete picture of what customers actually want to know. That is the basis of the whole process.
Once you know what customers are asking for, you can start the conversation. That is more than just formulating answers. ANWB chose to write its answers in a conversational style, in order to make the interaction between its customers and Iris as human as possible. Answers had to meet the following conditions:
• Find out what the customer wants
by asking further questions. Only if you really know what the customer wants, you can help him / her further.
• Show that you understand the question
instead of immediately dropping in. You can take it a step further by incorporating positive confirmation in your answer. For example, if a customer asks whether the car insurance covers windshield damage, you could answer with, “View the conditions of our car insurance”, but you could also say: “It's good to find out what is and what is not covered! In My ANWB… ”. Sounds better, right?
• Provide information in small doses
Only provide the information that is relevant and try not to summarise all the information on a particular topic in one answer. Avoid overloading the customer with information. The same goes for links in answers: only offer them if they are of added value.
The most important lesson is: you are never done. A chatbot and the content it contains are continuously subject to change. "We started with sixteen conversational dialogues, but it didn't end there," says Margot. “You have to monitor continuously and make adjustments where necessary.” Most importantly, you have to be able to respond quickly to customer behaviour. Carolina adds: “It also helps that a large group of colleagues was involved from the start. This allowed us to make changes in the content quickly. ”
After analysing the data from the holiday preparation period, this approach proved to be a great success! The analysis showed that Iris was better able to answer questions and more customers were helped. Carolina: “The KPI we use to measure whether Iris answers questions correctly is the negative feedback. We never managed to get it below 10%… until the summer of 2019. ” In this period, the negative feedback almost halved and sank to 6%. In addition, Iris also answered twice as many questions about holiday preparation as in previous years. ”
Select a region to show relevant information. This may change the language.
Is this region a better fit for you?