Services with artificial intelligence

Does every company need a chatbot?

Whether for flight reservations, in customer services or online shopping: Swiss companies are also making increasing usage of chatbots. Where they are failing and how you can get your own chatbot.

Text: Adrienne Fichter,

We all know the situation: Minute after minute spent on hold listening to music on a hotline, with no end in sight. Or we call a service number at 5.35 p.m. at the very moment the customer advisor is leaving work. In the future we won’t have this problem. And we’ll be able to communicate with the company whenever we like, even at 3.00 in the morning.

The software firm Oracle predicts that customer service will undergo a complete overhaul. The research firm Gartner forecasts that barely a third of customer relationships will require human interaction. The flipside of this: two thirds will be maintained by automated software.

Will the personal call centre agent soon be a thing of the past?

Chatbots are taking a lead here. Chatbots are chat robots, or smart programs that can communicate autonomously with users. They can learn and are programmed to receive data and react to specific keywords and phrases. This enables, for example, shopping and service processes to be supported based on a dialogue. Of the sales and marketing managers surveyed by Oracle, 80 per cent stated that they want to implement this technology by 2020. Wide-ranging messenger portals are an especially popular setting for the interaction.

For example, within the WeChat app – the Chinese equivalent of Whatsapp – 650 million customers make purchases, orders or reservations every day. This Asian all-rounder app is now being used for practically everything, from shopping to online banking to ordering a taxi. In the West, on the other hand, it’s only a matter of time before Facebook dominates the messaging platforms, according to Jessica Ekholm, Research Director at American market research firm Gartner.

Facebook Messenger not only has its large potential target audience to offer. It’s also the case than many companies already have a corresponding fan page that can be extended via a simple modular system in order to develop a bot. At this point 34,000 bots are looking after users at Facebook Messenger for a wide range of companies. However, in contrast to WeChat, they are used “only” for customer service, recommendation marketing or competitions. Some well-known German-language examples: Chatbot Mildred helps Lufthansa customers to find suitable air fares. The chatbots of H&M and Zalando give friendly shopping advice. The Botlist website provides a good overview of the existing chatbots, their digital “catchment areas” and the specific service they provide.

One-stop shop for AI

The Swisscom competence centre for applied artificial intelligence offers companies everything they need for the quick, successful implementation of projects in all areas of artificial intelligence, from consulting to the right technology to integration.

To the offer

“Customers want to get information as quickly as possible, and a chatbot may be the answer – or it may not.”

Marc Steffen,
Head of Product Design Artificial Intelligence & Machine Learning Group, Swisscom

Where does Switzerland stand?

In Switzerland many big companies from the retail specialist, insurance or banking sectors are following the developments for now while working away on their own projects. However, most of them are still at the testing stage or are using the technology merely to improve the administration in the background for the time being. Swisscom subsidiary is experimenting with three chatbots in Facebook Messenger. They can be used to call up the cinema listings, the weather report, the train timetable. Since last December, telecommunications provider “Wingo” has been using bot “Cara” on Facebook. This is especially practical for customers who want to talk to somebody at Wingo between 6.00 p.m. and 9.00 a.m.

The challenges

Just how advanced is artificial intelligence technology at this point? Is it a valid substitute for human advising skills? Many test reports show that only a few providers make good on their promises in the Facebook Messenger app. Experience has been sobering, as a self-test by Zeit Online has shown: Even in the realm of its supposed core skills – the weather report – weather chatbot Poncho fails occasionally. The verdict of the media portal is that entering “Weather Berlin” on Google would be more efficient. Poncho provides this information in a rudimentary fashion at the end of a two-minute dialogue.

Marc Steffen, Head of Product Design, Artificial Intelligence & Machine Learning Group at Swisscom, warns about false expectations. Chatbots will not always provide the right technological solution for a company: “Customers want to get information as quickly as possible, and depending on the target group and the context, a chatbot may be the answer – or it may not. Maybe the required information can be found more quickly via a couple of clicks on a regular website than via a bot.” Companies should always ask themselves: “Would a person be better for the end user?” If so, a chatbot is not needed. Bots shouldn’t be trying to replace people in positions where people fare better. Rather they should be used where people are too slow, for example.

Swisscom and EFPL are working on artificial intelligence

Most chatbots score highest with simple questions. Here they guarantee savings for the company in terms of resources, money and time. According to Steffen, in technical lingo they’re known as “retrieval bots”. In collaboration with ETH Lausanne (EPFL), Swisscom has founded a competence centre to enable the use of artificial intelligence for Swiss companies quickly and easily. Various types of chatbots are being researched, from simple “retrieval bots” to personalised chatbots. Machine learning is central to the development of sophisticated chatbots. “The chatbot is always only as good as the available knowledge,” says Steffen. The catalogue of Wingo chatbot Cara comprises 25 simple questions. Cara deals with 20-40 requests a month. Its functions still have room for improvement. At present Cara does not deploy artificial intelligence and therefore cannot learn autonomously. But this is going to change soon. Because the more commands and data that are fed in, the more specific and personalised Cara’s responses will be in the future.

“The chatbot is always only as good as the available knowledge.”

Marc Steffen

How does one build a chatbot?

How does a company go about getting its own chatbot? On the one hand, platform operators Google, Facebook and Microsoft offer assistance. The two biggest rivals here are Microsoft and Facebook, with their very different development environments. Facebook Messenger, for example, enables developers to design little assistants with its own open Messenger platform. Microsoft has come up with the Cortana Intelligence Suite. Developers and companies can use the Bot Framework and the Azure Bot Service from Microsoft to program self-learning applications and bots. Additionally, various ICT providers and agents in Switzerland have specialised in this area.

Are chatbots just a passing fad, marketing hype? Or a serious phenomenon for companies? They could be the start of a company-wide mobile strategy. Or they could be used as a temporary solution en route to intelligent assistance systems like the personal digital assistants we know from Siri (Apple) and Allo (Google). In any case, chatbots mark the start of an exciting journey into the world of artificial intelligence. The question as to whether this is worth investing in should not depend primarily on a company’s resources, but rather on whether the customer really enjoys added value through communication with a machine.