Swiss companies’ chatbots suck – but that should change

Their names are Ollie, Nelly, Clara or something like that. Some introduce themselves. “Hello, I’m sam!” Or: “Hello! I’m Laura.” Others are less brash and wait to be contacted. All are helpful. They ask, “What can I do for you?” Or say, “I’m happy to answer your questions.”

You come across them while surfing the websites of hotels, insurance companies or banks. Chatbots are everywhere. Hardly any larger company can do without a digital assistant on its website.

They’ve been around for a while, but since ChatGPT catapulted itself into the collective consciousness of users, they’ve taken on a new meaning.

Chatbots are no longer just annoying windows that appear on a website, but can also be almost omniscient interlocutors. Unfortunately, the latter rarely applies to the Swiss copies of the digital helpers.

Where’s the problem?

“Many chatbots are really just dummies. They just forward the users to customer service,” says Jörg Neumann, director of the consulting firm Neumann Zanetti & Partner. With his team, he conducted one of the first market studies into the use of chatbots in Switzerland.

The conclusion: Most users are open to this type of communication, but are quickly disappointed when it turns out that the bot cannot satisfactorily answer the questions asked.

Neuman says:

Things get particularly grotesque when a bank bot doesn’t understand the word “bill”, an airline’s digital interlocutor doesn’t know what “last minute” means or a public transport company’s voice robot only understands “super ticket” at the station. These are all examples that survey participants criticized.

Why are chatbots so bad?

Sophie Hundred marks conducts research at the Lucerne University of Applied Sciences and advises companies on the use of chatbots. She says, “The chatbots often fall short of users’ expectations.”

This is not because of the technology itself, but because of how the companies implement it. “The best worker in the world can’t be good if he’s not trained properly,” she says.

Ideally, a chatbot takes over the simple tasks of a receptionist or receptionist. For example, he can help a hotel guest to postpone his booking, support an insurance customer in reporting a claim or hand over the necessary documents to a resident of a municipality.

But you have to train them for that. Simply put, there are two different types of chatbots: the dumb and the adaptive. For the stupid you have to formulate an appropriate answer to every expected question. For this, important words must be marked manually. For example, if a hotel guest uses the verb “open” and the noun “swimming pool” in a question, the chatbot will reply with the opening hours of the swimming pool – and maybe even the fitness center afterwards.

When the same questions are asked over and over again, such as on a hotel website, this technology can go a long way, says Hundertmark.

But the future belongs to adaptive chatbots such as ChatGPT from the American company Open AI. Specific answers are not given here for questions. Instead, the chatbot generates its own answers. For this purpose, a neural network is trained with as much data as possible.

How ChatGPT finds its way to Switzerland

The insurance company Helvetia is probably the first Swiss company to use this technology. It uses ChatGPT’s artificial intelligence and language understanding and trains it with the full knowledge of its databases. In this way, the chatbot must gain a detailed understanding of the insurance industry and be able to answer tricky questions accurately.

Sophie Hundertmark from the Lucerne University of Applied Sciences and Arts supports Helvetia in setting up her new voice robot Clara and scientifically evaluates the project. The insurance company calls the chatbot an “experiment” in which everyone can help with their questions to improve it.

Hundertmark would like to get more Swiss companies excited about ChatGPT, but she feels a lot of reluctance. “A lot of people just want to wait and see,” she says.

This is probably also because the language robot has made some negative headlines since it started taking the internet by storm in November. Thousands of researchers and experts issued a letter warning against artificial intelligence and calling for a six-month moratorium. ChatGPT continued to stand out with almost grotesquely wrong answers. And in Italy, ChatGPT’s service has been suspended due to privacy concerns.

According to Hundertmark, consumers are much less afraid of data protection and the like than is currently conveyed by the media. She is doing research and has evaluated the first – yet unrepresentative – results. The conclusion: “Clients would appreciate businesses using ChatGPT, especially for general inquiries and initial consultations.”

And now you!

What practical experiences do you have with chatbots in customer service? Is the chatbot consultant right when she says consumers are not concerned about privacy and security?

Write to us via the comment function!

Sources

(aargauerzeitung.ch)