By Dr. Sara Detzler and Luise Hogrefe, Presales Specialist for SAP Conversational AI.
Hi there. I can make your interaction with company software more intuitive.
Suddenly these things called chatbots were popping up (and even recently generated a huge hype). At first that puzzled us. We thought mankind invented the internet so that they have to talk to each other less? And now everyone wants to talk with bots and—even less understandable—they even want to talk with their SAP S/4HANA systems? But when you think about it, it makes perfect sense.
For example, remember old SAP systems transaction codes like /nstfk, which are far away from being intuitive? We still get nostalgic thinking about our little book where we wrote down the codes like vocabularies. When we were in university and we had to work with a bunch of transaction codes, we joked that SAP stands for “Schreibs Auf Papier,” which is German for “write it down on a paper.”
But let’s face it, the pure need for such a book gets you thinking. Wouldn’t it be great if you could just ask a bot for what you want instead of looking in your little notebook for one single code? Our mother language is the first way we learn to properly express our wishes and needs. So why not use this most natural and intuitive way of interaction to make business software more fun to use? In addition, our world is becoming more and more conversational/viral, thanks to social media. Thus, it makes perfect sense that business software follows this trend.
You see, for business software the decision is quite simple. Just try it, there is a high chance you gain benefits. When it comes to communicating with your customer, it’s perfectly understandable if you have some concerns.
Hi there. I can improve the experience for your customers.
The fact that chatbots bring measurable added value is widely known and accepted. Compared to humans, they scale better, they are always on duty, and speak multiple languages.
Sorry to say, but thinking about certain finance sectors, their opening hours are not really customer friendly, especially when you work a regular job. Luckily, chatbots don’t come per default with opening hours.
Another drawback of humans we experienced lately is that they can be impolite and incompetent. And not just that, we don’t even want to count the amount of service agents we have to go through, that tell us “Let me forward you to XYZ,” until there’s a finally an agent we can tell the customer ID to.
In one recent case, the responsible service agent didn’t let me (Sarah) finish describing the problem, was incredibly rude, and in the end gave me a completely wrong answer. After I questioned the correctness of his advice and requested to check the solution together, the agent got even unfriendlier and treated me like a really stupid five-year-old girl. The first thought that came into my mind was “How could a chatbot have done much more wrong?”
For us, these kinds of examples are the best motivation to promote chatbots. Don’t get us wrong, there are a lot of great service agents, but let’s face it—customer service can’t keep up with all the incoming requests from customers anymore.
Furthermore, we have to admit we are the type of customers who read everything before signing it and ask quite detailed questions if we don’t understand something. Maybe the appearance of my (Sarah’s) nerdy comic shirt in combination with the difficult questions confuses the agents, but still people like us can be rewarding customers too.
So, how could a chatbot help to improve the status quo?
First, chatbots don’t care for your appearance, gender, or anything else.
- They treat everyone equally and as friendly as you programmed them. Of course, you could still program them grumpy or unfriendly if that is part of the companies’ strategy. (We don’t see why someone would do that but the strategy “They can afford to be unfriendly to their potential costumers they must be doing a good job” might work. Couldn’t find any reliable data to investigate.) Maybe they don’t even need to be unfriendly, but you can definitely add some funny jokes or sarcastic comments because we don’t need data to know that everyone loves a Little humor.
Besides, a chatbot always gives the answer that is stored in the backend it’s connected to.
- Therefore, it is possible to quickly adapt bots to changes or new regulations, by changing the knowledge base in the backend. A nice side effect is that chatbots don’t have the urge to convince their counterpart of their competency. Of course, everyone values competent support, but the thing is, showing off with expertise and really trying to help someone are not always the same thing.
Most people think that communication has the goal to transport a message, but for us the chatbot’s main goal is to provoke a reaction from their discussion partner.
- This way of thinking helps us to shift the focus completely from yourself and your message to the other actors in your conversation. In contrast, if you concentrate on showing your own competency, your focus will simply not be on the customer and his problem and he won’t feel supported.
Chatbots are usually programmed in a way that they listen and ask for all information they need to answer a request.
- If they don’t know the answer they don’t get rude or try to convince the customer that he asked a very stupid question; unless explicitly programmed. They are not ashamed to say that they don’t know the answer to a problem and will refer you to a specialist who can help you.
Hi there. I love to cooperate with humans to learn and improve my skills.
This leads us to the topic of supervised chatbots. It is a great practice to combine the best of both worlds, chatbots and good service agents. The chatbots can solve the easy and simple standard tasks, while the service agents can supervise them and improve their capabilities, taking into account the questions they can’t answer. Furthermore, the service agent can take over the conversations if the bot gets stuck.
So, it looks like bots and agents can complement each other quite well. As mentioned before, agents need to make sure that they focus on the customer and make them feel comfortable. How can we motivate a machine to do so as well? Here, the right modeling of the bot is important.
First, we define how the bot should interact with customers. This starts with small things, like how to introduce itself, asking if a short or long answer, or if the bot should use emojis. Other thing we always train our bots for are greetings, saying goodbye, and understanding expressions of laughter. For some customers, it is also important that the bot can tell jokes, talk about the weather, and answer the question to the meaning of life. For mysterious reasons these are the first things users ask for when they talk to a bot.
Probably they sometimes enjoy playing with them and pushing them to their limits. Or, since humanity has been trying to answer the question for the meaning of life for decades, they’re just curios if a bot has the ultimate answer. These things that make the bot more fun to use, are just as important as the competency to solve the tasks itself.
That is why when building a bot, always take in consideration that it’s learning from how you use it. Thus, sufficient time to really test it is so important. For us, this is a lot of fun. We always think of them as “our cute little bots” and it’s great to see how quickly they learn and evolve— reminds us of little kids when they started talking.
We can just recommend you try it on your own. It might be easier than you think.