Patrick Daxenbichler - Fotolia
An increasing number of Dutch organisations are using chatbots to get in contact with potential customers. Machine learning and artificial intelligence (AI) play a part in the innovation, but economics is the main driver.
One institution that has experimented with a chatbot is the Dutch police force. The organisation recently started work on a chatbot of its own, called Wout. The name is a play on words, being both a common Dutch first name as well as slang for police officers.
Wout can receive common complaints from citizens about things such as noise disturbances, complaints that currently can only be reported by phone.
After choosing from a set of possible predetermined situations, Wout asks users to give more details about their complaint. It asks questions like where the disturbance is or how many people are involved.
Currently, Wout only asks its users closed questions, all of which are carefully mapped out into a decision tree by developers.
In the future, the police service wants to make it possible for users to ask their own open-ended questions, project leader Bert Visser told newspaper Tubantia.
But making that change is not easy, as it requires a lot of machine learning – something that can become a complicated and expensive development.
Preprogrammed or artificially intelligent
Chatbots come in two forms. The most common are chatbots that work based on predetermined, programmed responses – but while this particular form is the easiest (and cheapest) to build, a lot of companies are opting for smart, AI-enabled, machine learning bots that form natural, flowing sentences and can understand humans. Of course, such chatbots are not straightforward to create.
There are however some large companies in the Netherlands with the know-how and money to build advanced chatbots, and perhaps the most prominent example is BB, or BlueBot, the adaptive smart assistant built by airline Air France-KLM.
BB lives on Facebook Messenger, where it helps travellers find and book flights, as well as remind them to check in. BB is connected to KLM’s API’s – but also to the company’s Salesforce CRM system so that webcare employees can easily take over the conversation when BB doesn’t know what to say.
Read more about chatbots and AI assistants
- Suppliers claim consumers want self-service chatbots. Turns out they'd rather scrub toilets or go to the dentist than converse with customer service in general, bots included.
- One Nordic bank decides to stop using cognitive agent Amelia to focus on other AI technologies, but another widens the use of the software.
- Chatbots, software that uses normal language to communicate with people through text or voice, are proving their value in a surprising variety of ways – from accountancy support to mental health.
For KLM, a smart chatbot isn’t just a gimmick. In fact, since the company implemented BB, it saw engagement grow 40%. Customers threw a whopping 1.4 million queries towards BB, and 15% of all boarding passes are now sent to customers over Messenger.
There were two important reasons for KLM to start the implementation of chatbots. One was the increasing demand from customers on the quality of their conversations. Particularly important is response time, said Martine van der Lee, director of social media at the airline giant. “Customers expect instant answers to questions.
“Their expectations have grown tremendously over the past two or three years, particularly regarding response times. The customer conversation has become much more real-time, and it’s up to us to find a way to address that.”
Of course, there’s also an economic reason. As the number of questions from customers grows, so will KLM’s social media team. Chatbot BB can handle all those extra requests for a fraction of the costs. Currently, the webcare team has over 200 people working for it. “It’s not feasible to keep that number growing,” said Van der Lee.
From dumb to smart
While building a simple chatbot with preprogrammed answers can be cheaper, they can eventually become more advanced, utilising all the collected data. This is the strategy that insurer Reaal uses.
“Our first bot was programmed with the answers to 70 of the most frequently asked questions,” said Constant Moolenaar, director of customer services.
“We built the chatbot in four weeks and put it live to gather as much data as possible. Content is added in the form of answers from customers. The next step is finding out what questions are and aren’t answered. That forces us to constantly optimise the process.”
Machine learning can eventually transform a simple chatbot into a more advanced chatbot, but this is where Dutch language proves a barrier. With only 28 million native Dutch speakers worldwide, providers of AI software don’t tend to focus on the region. That can make it difficult to use platforms mostly focused on the international market.
Natural Dutch Language Processing
Dutch startup Embot aims to change this by building a chatbot specifically focused on the Dutch language. It uses Natural Language Processing to analyse the responses of chatbots users to grow its understanding of the language.
The first application to officially use Embot is Bag To School, a Facebook Messenger bot that helps students pick out backpacks. “Based on the product data for all available bags, the bot learns what colours are available,” said Embot CTO Serge Cornelissen.
“The chatbot uses machine learning to understand what’s important when picking out a bag. If you want a waterproof bag in black, you can ask the bot for that.”
The chatbots used in the Netherlands are, just as internationally used chatbots, built on common platforms that plenty of people use. “Facebook Messenger is installed on seven million phones in the Netherlands,” said Tom Huijgen, data strategist at the company behind Bag To School. “It’s the platform people use to talk about new products.”
KLM have the same reason for using Facebook Messenger. “We want to be where our customers are,” said Van der Lee.
Customers, of course, are mostly on social media, and Facebook Messenger is the most popular platform. It’s used by not only KLM but also hundreds of others companies in the transport industry or the medical sector. Chat apps like Kik and Telegram are less popular, in line with general use in the Netherlands.
Notable is how many websites and webshops still utilise their own platforms for chatbots rather than external ones like Facebook. The country’s largest webshops Coolblue and Bol.com both built chatbots for their websites, helping customers pick out products get in touch with their customer service. This is also the case for popular second-hand marketplace Marktplaats.nl.
Chatbots still have a long way to go – especially in the Netherlands. But most companies at least seem to understand the possibilities of them in their customer service processes.