Susanne Lebkücher works at Deutsche Telekom at the interface between human and artificial intelligence (AI). She tests new developments such as digital assistants directly with customers, in workshops for instance. “Users want to know what happens with their data,” she says in the interview.
With Siri and Alexa, digital assistants have long become part of the daily lives of many people. How do they interact with these innovations?
Susanne Lebkücher: Around three years ago, consumer’s expectations of these kind of assistants were low. Customers knew a few of them, but did not use them often. Take for instance our chatbot Tinka, which we have on T-Mobile Austria’s website. Many saw the bot as a classic search function. Or simply as a way of accessing the live chat with an advisor. Nowadays lots of people are familiar with the keyword “AI.” Even so, it is still tends to be just some vague idea or associated with science fiction. Yet we are perceiving a surprisingly positive mood overall. There is a growing openness particularly toward Siri, Alexa and the Google Assistant, while the requirements placed on AI are also becoming more demanding.
You are testing these kinds of digital assistants or chatbots intensively with customers for Deutsche Telekom. What do people want to see?
Susanne Lebkücher: Simplicity, help with simpler everyday problems. And trying to see the bigger picture – for instance when a chatbot identifies network faults in a certain area. In short, whenever it is easy and rapid, and provides practical added value, consumers tend to turn to these kinds of solutions.
If you transfer this knowledge to Deutsche Telekom digital assistants – what is the recipe for success?
Susanne Lebkücher: The assistant must be able to recognize what the customer says. We do our utmost to ensure our chatbots understand natural speech patterns. So even when someone describes their problem using their own words. The answers should be clear and structured. The system also needs to be able to summarize conversations. Customers want to be able to steer the dialog. The assistant should also be familiar with the customer’s context, i.e., such as their history with us, and be able to respond accordingly. In this respect, access to personal data is also important. And not least, it is about simple use scenarios that work well.
For instance?
Susanne Lebkücher: If the Internet does not work, the person affected would prefer to speak to customer service experts over the phone. But if someone is interested in mobile rate plans, they will normally spend more time. So they will tend to use a tablet or PC and be more willing to use the chat function. The less complex and urgent the problem is, the more welcome digital assistants are. That is why we are starting with more basic applications to build up trust. Once customers learn from experience that it works, they will also ask the chatbot about more difficult topics. It is important to us that users are not left stranded and if necessary they can be forwarded to a human advisor.
Keyword trust: At Deutsche Telekom you work on the premise that a digital assistant presents itself to the customer as what it is.
Susanne Lebkücher: Solutions such as “Google Duplex” provide a taster of the future of AI. Google Duplex is a natural-sounding bot that makes phone calls which are or should be almost identical to a real human. But the trick for us is that the assistant does not claim to be something it is not. It should show understanding without pretending to display empathy. “I can understand your frustration” is not acceptable – whereas “I can help you with this” is. One customer requirement which we repeatedly hear in our conversations is particularly important to us: Users should be clear from the outset who they are dealing with. Deutsche Telekom has also set that down in one of its nine guidelines for artificial intelligence. Transparency is a huge issue for consumers and for us.
And how important is it to customers that their data is secure?
Susanne Lebkücher: Participants in our studies constantly ask us: “What happens to my data?” And for that reason too the aspect of transparency is firmly incorporated in our guidelines. AI algorithms and functions may well be a mystery to customers, but those same customers are generally curious to find out more. They want to know how and why the assistant comes up with an answer. In this respect, there is a fine line between protecting and releasing your own data. Customers release their data as soon as an application seems practical to them and delivers added value. There is often no understanding about how their data is processed. Deutsche Telekom is also responsible here for informing and explaining these aspects to people.
In your experience, what is important for companies that develop AI?
Susanne Lebkücher: Having AI experts and specialist knowledge on board. The team should be international, include women and men as well as different viewpoints and attitudes. We are in the midst of a learning process which we can shape. But only together with customers and employees. AI can relieve us of “tiresome” tasks and give us the opportunity to focus on what creates value. But first you need to recognize where it can deliver added value. What are the relevant needs? How do consumers want to be addressed by AI? What tasks can people perform better? And you need to closely integrate future users to find that out. Otherwise, you create powerful systems that do not solve any real problems.
Developments with customer feedback
In an expert team, Susanne Lebkücher tests AI developments from Deutsche Telekom directly with customers - before, during and after the launch of these innovations. It is about chatbots and voice assistants on the phone just as much as it is about applications that improve workflows within the Deutsche Telekom Group. The team starts with observational studies, accompanying prospective users in their everyday lives to determine their needs, problems, and tasks. Then it collates the results, analyses them closely to understand customer preferences and behavior. It gets hold of additional expert opinions. Finally it stipulates which customer problem it aims to solve first and what applications are particularly suitable. Afterwards, further development takes place in an interdisciplinary project team. Meanwhile the tests with potential users continue in workshops. That helps determine early on whether everything is going in the right direction, whether ideas need to be modified or discarded altogether.
Examples of chatbots at Deutsche Telekom: Digital service assistant at Telekom Deutschland, Tinka at T-Mobile Austria and Sophie at Congstar.
