AI and chatbots create the customer communication of the future

They can send text messages, order a taxi and keep you up to date with the latest news. The market for customer service bots and virtual assistants has grown rapidly over the past year. Programmer Giovanni Toschi, conversation designer Ditte Wolff-Jacobsen and economist Stefan Fölster believe that this trend has only just begun.

Robotar-875x580.png

Hi there, sleepy head! The weather today is clear with temperatures between 2 and 8 degrees. An ideal day to go out and...sit in front of the computer. No matter how hard I try, I can’t think of anything fun to do outdoors."

The weather bot Poncho wakes me up with today’s weather forecast and a never-ending flow of GIFs and bad jokes. 

Its simplicity, value and the personal touch are presumably some of the reasons for its popularity. Facebook launched support for chatbots in Messenger in spring 2016, and there are now just over 100,000 bots on the platform.  

The art of designing a conversation

Ditte Wolff-Jacobsen runs an agency called iindhold in Copenhagen. For a little over twelve months, she has been working almost exclusively on conversation design and digital content. Interest in adding a chatbot to websites and social channels is growing rapidly. 

"It’s a good way of switching from informative texts to a conversation," she explains. 

A chatbot can answer standard questions about products and services around the clock while collecting data from every single interaction. Messaging platforms also offer an opportunity for continuous follow-up. When the customer is close to a shop that was visited previously, a bot can offer a discount coupon to take along or provide information about new products in store. The result is long-term relationship-building in which customer service and marketing merge and opportunities increase to find out more about what customers want. And just as with Poncho, it’s possible to communicate personality and tone of voice that are line with the company’s values. 

When Ditte Wolff-Jacobsen is asked to develop a chatbot, she starts by confirming the benefit and the expectations. 

"The first thing you need to do is to identify the right location for the interaction," explains Ditte Wolff-Jacobsen.

Ditte Wolff-Jacobsen runs an agency called iindhold just a stone’s throw from Torvehallerne in Copenhagen. Customer service often has the answers

One good place to start is to analyse your data and talk to those employees who have personal contact with customers. Which questions keep on cropping up? Which routine tasks would it be helpful to have automated so that they have more time for other things? Once this has been done, it’s easier to decide where the needs exist in the business.

There is then a brand discussion to define the tone of voice and personality, before she designs the actual content. 

"Once we have this in place, we produce a prototype and see what happens," says Ditte Wolff-Jacobsen. 

When a customer encounters a bot, there are often pretty high expectations of what it can achieve. But at the same time the technology is new, and a learning processes is required for both customer and user regarding which questions a bot is able to answer. As a general rule, you can say that it’s better to create a bot that knows its subject than one that has witty answers about this and that. 

If you expect to be able to build something that can answer everything, you’ll be disappointed. But most people think that this is an enjoyable format and intuitively they want to start interacting. 

The design of the bot can be anything from the simplest possible dialogue boxes with predefined response options to advanced AI that understands language, books meetings and gathers data for the company’s CCM system. 

If the bot is to work optimally, you must regularly evaluate, capture errors and correct them.

"You shouldn’t view it as a product or a channel, but rather as a new colleague. If you don’t want to teach the colleague to be a skilled customer service employee, you shouldn’t have it."

Chatbot-startup attracts big clients

Giovanni Toschi is one of the founders of Danish company BotSupply. Since 2016, they have been building chatbots for more than 20 Danish brands, including the insurance and pension company Top Denmark, Danske Spil and Dansk Supermarket Group. It was when they saw the rapid development of AI that they decided to invest.  

"We’re now a startup with a network of 15 people working for us," explains Giovanni Toschi. 

Customers have paid anything from SEK 30,000 to SEK 300,000 for the development of a bot. The difference in price reflects how intelligent it should be and whether it has to be able to interact with the company’s other systems. If the bot has to learn how to understand language, they need to link in artificial intelligence (AI).  

Put simply, AI sorts how words belong together. The word ‘man’, for example, relates more closely to the word ‘king’ than ‘queen’. By systematically analysing language, AI decides how close together the words are in their meaning. 

"If the bot is to be able to discuss specific areas, it must be programmed for this in a different way than for general questions. It might be, for example, a bot that knows everything about a range of furniture," explains Giovanni Toschi. 

Italian Giovanni Toschi is one of the founders of the AI company BotSupply. He made his way to Denmark as a student when he realised that technical developments were moving faster there than in Southern Europe. A multilingual colleague

With AI, it’s also possible to have customer service in several different languages. Right now BotSupply is working with a developer in Dubai who wants to build AI for Arabic-speaking users. 

"They can use our system to educate the language engine, and in return we get a new language."

As a startup, it is precisely this joy of innovation that is their primary competitive tool. At the moment they are unable to offer employees in the network high wages, but they do have the freedom to work on the thing about which they are most passionate. 

As with all new technology, sometimes things go better, sometimes less so. One example of a bot that really had a bad time was Microsoft’s TayTweets, which had to be shut down after 16 hours when it was transformed into an Internet troll. 

"A bot is like a child. If you communicate with unpleasant people, you become unpleasant. Tay met users who swore and told racist jokes," says Giovanni Toschi. 

Not like in films

In science fiction films it’s not unusual to have highly intelligent AI units threatening to take over the world, but in reality it’s the human influence that many experts believe we need to be most aware of. 

 The Swedish economist Stefan Fölster has just issued a book entitled ‘Robotrevolutionen’ (‘The Robot Revolution’). He describes the use of robots and AI as a shift in society with massive consequences.

"Ultimately there are existential issues. In contrast to previous technological shifts, which resulted in a high standard of living, we’re now dealing with the question of what makes us humans unique. Are we needed?"

This question is rapidly becoming increasingly relevant in the workplace. As they become smarter and learn how to ask follow-on questions, they will take over more of the interaction with customers. Robots will be able to do most of what we can do, even better than we can ourselves. But services need to be flexible so that it’s possible to switch quickly to a human employee when the need arises. 

"There will be customers who never want to speak to robots, but many will discover the benefits."

Stefan Fölster is Head of the Reform Institute in Stockholm and author of the book ‘Robotrevolutionen’ (‘The Robot Revolution’). Keeping track of data

A robot can feel more reliable, as it has all the latest information and treats everyone the same. 

Swedish bank SEB is one of many major companies that have started to experiment with digital systems. Their customer chat function informs visitors that they will be assisted by SEB’s digital customer service employee, Aida. It’s in the area of AI where technical progress is moving most quickly, and the banking sector is believed to be one of the areas where it will quickly make a major breakthrough. Not just in customer service. A robot such as Aida affects the transfer of knowledge within the company. The bot can train new employees and human employees can in turn train digital assistants. 

"This will also be generating a lot of discussions about integrity in the future," says Stefan Fölster. 

When the bot sees that you are doing something you should not, it can ask whether you want it to ask the manager for permission. This can of course affect the relationship between employer and employee, but also between customer and company, where there is a major need for more modern personal data legislation.

"GDPR will definitely not be the end of the story. My gut feeling is that there may be such major benefits in storing personal data that the law will become more differentiated."

He cites the two examples of healthcare and crime-fighting, where the greater storage of personal data can bring major benefits. IBM’s supercomputer Watson has shown in tests that it can make better diagnoses than doctors. 

"As diagnostic systems develop, it will be unreasonable to resist." 

The breaking point is close

AI is described as the ‘electricity of our age’, and those companies that use it smartly will enjoy major advantages. For private individuals it will also be an indispensable way of finding relief in a stressful everyday life. For the time being, Stefan Fölster describes chatbots and digital assistants for private use as being fun but a little cumbersome, but we will soon reach a breaking point. 

"Once this has been reached, they will spread extremely quickly. Just like mobile apps. Initially they were toys, but then enormous value appeared with services such as mobile banking and bank ID."

For a long time, Apple’s Siri led the way among voice-controlled assistants, but competition is now on the increase from many sources. Google Assistant, for example, has developed a memory and can understand references to information mentioned earlier during a conversation. One clear trend is that we are moving from screen to voice. 

"It’s good not to be walking around staring down at a screen," says Stefan Fölster. 

As usual, children and young people are first to adopt the technology, and the adults then follow. Children who grew up with a smartphone in their hand don’t hesitate to ask the voice assistant to play Lady Gaga’s new single or to read a goodnight story about the robot chicken who sleeps under a banana leaf (say: ‘Siri tell me a goodnight story’ if you want to hear it).