"I want to talk to your chatbot!"

Everybody talks about chatbots, but far too few talk to them. This is the only way to learn what they can and cannot do.

My enthusiasm for chatbots stems from the time when I watched every episode of the science fiction series "Star Trek: The Next Generation" week after week. There are basically two chat bots - not that I could have called them that at that time 🙂

One is known to most as Lieutenant Commander Data, the android with the silver skin. He symbolizes what I hope for from artificial intelligence and chatbots: High communication competence paired with a keen interest in us humans.

The other chatbot is the starship Enterprise itself. The on-board computer, although perceived to be larger than Commander Data and omnipresent, seems less smart to me, but instead performs clear tasks such as retrieving information, locating people or ordering tea from Captain Picard: "Computer, tea, Earl Grey, hot. 😉

I'll just call Data, without wanting to offend him, the marketing version of a chatbot. But while Data has many fans, the on-board computer has none - although it usually does its job more reliably. For me it is the UX version of a chatbot.

Data or board computer?

And exactly between these two poles of Lieutenant Commander Data and the Enterprise's on-board computer is the term bubble around chatbots. As early as 1966, ELIZA, the first chatbot, demonstrated the possibilities of communicating with machines using natural language. And already 50 years ago, it became apparent that humans had no problem at all with "talking" to this machine.

However, it soon became clear that nobody really needed this general talking machine because, like Data, it has no use case. Artificial intelligence is still so much in its infancy that, like this Eliza replica, it has so far only feigned understanding. That may change one day - but unfortunately we are still far from that.

But that does not matter. You don't need real artificial intelligence to develop meaningful chatbots. Nobody needs a whole data set for customer service - in most cases, a little onboard computer would be enough. But where exactly is the difference?

Chatbots need a Use Case

Eliza's weakness is that as a chatbot she has no concrete task. The on-board computer, on the other hand, solves all the problems that computers on a spaceship have to solve. So as soon as chatbots are developed on the basis of concrete use cases, they can very well be useful. Data also has no concrete task - what gets him the many fans, however, is his striving to appear as human-like as possible when dealing with people. To achieve this, he imitates humor and similar qualities - sometimes more, sometimes less successful. 🙂

From it some can be derived for Chatbots at the current stage of development. For example, that even restricted programs can make a lot of sense, for example to strengthen the customer experience in customer support or to answer simple standard questions and thus save personnel. Chat bots can do this, because while the customer center cannot always be reached immediately by phone, bots can fulfill the initial request for contact and first answer immediately and at any time.

The answers of the chat bots do not have to be perfect in the first step ... but in view of the real-time expectations of today's users they have to come immediately. And if, like Data, they strive to imitate human manners without pretending to be human, then they will be accepted.

Chatbots move in a context

Today, many use cases consist of the query of an information or the selection from a variety of possibilities. Example: We start the weather app and select time and place. In the TV app, we select the film selection "Action" from the genres. We visit the airline website and book a flight from A to B on a specific date.

The information architecture and menu systems we use today are actually little more than a collection of ready-made answers to user questions. The art of UX is to answer the questions that are important in the context and actually asked by the user, and ideally not to bother the user with questions that do not arise at all.

Chatbots master this art almost by themselves. Because in natural language contact, no one would ever come up with the idea of reading out all available functions and menu items to the user. Instead, one feels one's way towards the goal in a dialog and asks if something is unclear.

Chatbots between sense and marketing gag

Anyone who has not yet tried out chat bots should do so - the existing offers already show that the clearer they dedicate themselves to a task and the more they emulate the on-board computer rather than Data, the better.

The paralegal RATIS is specialized in delays by airplane. With the help of a handful of questions, the chatbot interactively determines whether a right to compensation can be derived from a flight delay. In this way, it carries out important preparatory work, and at the end of the decision-making process, the chatbot can tell the user how much compensation he can expect - and, if necessary, connect him with the human lawyer.

Another example is Dinner Ideas: Here I get suggested dishes that I can cook with the ingredients and food I have in my fridge at the moment. Very useful and quite inspiring, even if the chatbot takes you to a different page when you want to get the instructions for a recipe.

Chatbots about everything?

Do we now have to switch everything to chat bots, and does everyone have to have one now? Of course not. If the problems to be solved are too general, the tasks too vague, then chatbots will fail - because we don't have any data yet. On the other hand, if the requirements are clearly defined, chat bots can do amazing things because they can concentrate entirely on the respective use case.

My favorite is Mildred, also known as Lufthansa Best Price, in my eyes one of the most convincing chat bots for Facebook Messenger at the moment. Not because Mildred would seem particularly imaginative and intelligent. It's because she does her very special job pretty well - picking out a flight for me in no time at all, faster than through the normal web interface or app.

Mildred neither tries to be an Artificial Intelligence, nor can it (currently) make complex entries. But she leads you to a first offer anytime and very quickly and is therefore another tool for customer retention.

Benutzerfreundliche Chatbots statt Hype

After a short hype in 2016, the chat bots were quickly criticized. They were too stupid, customers would reject them, not want to talk to machines and so on. Clearly, chat bots that act verbally intelligent, but are not willing or able to satisfy a single real use case, are ultimately perceived as a disappointment by the user.

The criticism may often be justified - but is often based on false expectations: Everyone hoped to talk to Data - but so far only one onboard computer has responded, and even that one is still clearly beta. And this will remain so for some time to come, because to want to use artificial intelligence via chat bots now would be a clear case of over-engineering.

That means: Good chatbots have a purpose and clearly defined use cases that serve it. And that also means: If you have limited, clearly definable use cases, it is worthwhile to deal with the topic of chatbots. Stores, delivery services, rail & cab, banking and insurance, even medical applications are conceivable where chatbots can query symptoms and thus make initial diagnoses.

For me typical cases for chatbots:

  • asynchronous requests from many users
  • requests at any time
  • questions with expectation of immediate answers
  • questions with only slightly personalized context
  • limited space for questions (at least at the beginning of a contact)
  • limited selection of possibilities how the end of communication can look like (answer; booking or product link; forwarding to the appropriate customer service area ...)

The good user experience with chat bots will then only depend on whether the user feels "understood" by the machine, i.e. whether his wish has been fulfilled. If the machine interlocutors accomplish this, no user will refuse to use their offer. And with the increasing availability of speech recognition on smartphones, chat bots will quite quickly become conversation partners with whom we communicate not by typing but by speaking. At the latest then it pays off to have already dealt with the topic dialogue via Chatbots.

In my next post I will give you 10 tips for good UX on chatbots.

Bettina Streit

Managing Director & Usability Engineer



Stay up to date!

With the newsletter, which is published 4 times a year, you will receive the latest news from our agency every 3 months. We will not use your e-mail address for any other purposes and you can easily unsubscribe from the newsletter at any time via a link in the mail footer.

Thank you! Your submission has been received!
Oh! Something didn't work. Please try again.