The Danish union Krifa found the right use of chatbots to maximize business value and beat the competition
Trade unions in Denmark: there are many, and potential members need tangible advantages to pick one over the other.
In such competitive space, it is crucial for trade unions to provide excellent customer service and support, whenever users need it, and to stay relevant with the latest technology.
For Krifa, one of the biggest Danish unions, a great customer service and a focus on happiness at work is the best way to attract new customers. This meant extending their support opening hours and providing faster resolutions. How to do this, without straining their human support?
It all started as a Proof of Concept to see what chatbots could do for their business and if it was even something that the members would use. Guess what? It turned out with a huge success rate.
Investment in resources
Read along as we have a chat with Benjamin Damsgaard, the Digital Consultant at Krifa, who together with his team takes extra good care of the chatbot ‘Kristina’. Throughout this blog post, you will discover how Krifa has taken advantage of all the opportunities that come with chatbots and how they have been experimenting with different bots to build the perfect chatbot for their members.
The making of a small scale chatbot
rifa is one of the biggest unions in Denmark and a digital-first mover. They decided to implement their first chatbot as a helping hand for their 200.000 members. But how has the journey been and when did it all start?
“We started in the summer of 2017 with a chatbot as a Proof of Concept to see, will it be something that our members will use? And we found out that it was!” – Benjamin Damsgaard, Digital Consultant at Krifa
It took Krifa 2-3 months to develop the content and set up the first bot. The bot was first built on a small scale as a proof of concept, to test how easy it would be to implement a chatbot and to get an understanding of the many usage possibilities. It was important for Benjamin and his team to create the bot fast so that the members could try it out as soon as possible. Their first bot was a customer service chatbot on Facebook Messenger, targeting the student members of Krifa, and could therefore answer questions that were relevant to them. They soon realized that the chatbot was a big hit:
“We had a huge success rate, the chatbot could answer many of the questions from our members. It was an area we wanted to pursue, to find out how we could use it for all our members.”
Reasons for implementing Conversational AI
Now that the chatbot had proven its worth and its future was secured, they had to make sure its aim and scope was clear to everybody, in and out the organization. Was the chatbot implemented just to save some money in the Customer Service department by removing human support? No!
“We hope with the chatbot that our employees in Customer Service will get better questions, so the questions that they are getting over and over again will be done by the chatbot. It’s not a way of saving a lot of money or to fire colleagues, but it’s a way of helping our members to get the FAQ easily. And to give them the answers even when we are out of our opening hours.”
Chatbots do have the power to reduce costs but in the case of Krifa, that’s just an extra bonus. According to Benjamin, it’s all about driving organizational efficiency and providing better customer service to the members as they are now able to reach Krifa 24/7:
“I think the business value for us right now is when we are helping people outside of our opening hours. So we can help them when our phone lines are closed.”
The ‘Chatbot First’ solution
Next step in the chatbot journey for Krifa was of course to explore new chatbot possibilities.A bot with a broader scope could potentially be of assistance to a lot more union members. Therefore, Krifa decided to focus on what we call their ‘Chatbot First’ solution: the bot was implemented on their website landing page. Here the members could get help and guidance via always-open, first-line support. The chatbot also had its own landing page. The chatbot provided guidance for all members that were affected by the new unemployment benefit rules and could, for instance, give answers to whether the user would be entitled to unemployment benefits or what action to take in order to be entitled.
“We wanted to give answers to all our members to help them in their work life, so we pushed a lot of information into the chatbot, (…) but it was too difficult to use for many of them.”
Despite good intentions, with the new bot, things seemed to become too complicated for the users. There was a lot of information stored, but finding the right information was difficult. Therefore, Krifa decided to rethink the bot and make another version.
The Christmas Calendar bot
Before we move on to talk about the newest chatbot version, we want to address another way that Krifa has experimented with chatbots to make use of their full potential. This happened in the winter of 2018, when Krifa made a Christmas Calendar for their members:
“Last Christmas, we decided to make a calendar in Messenger in our chatbot. So, we wanted to help people know what is Krifa all about. They got a question and then they could find the answer on Krifa.dk. So, the idea with the Christmas calendar was to help our members to know us. We gave them a link and a web page where they could find the answer. So, it was a way of showing people what we can do as a Trade Union.”
It was a daily chatbot calendar that throughout December gave the users advice and new input on how to improve their satisfaction at work (also called “happiness at work” or “arbejdsglæde” in Danish). For instance, if the user wanted to strengthen their motivation in their job search or at their job, the chatbot provided different articles written by Krifa on how to get the best results. This was thought through as a bonus to their members, but it was also a strategy used to spread the word about Krifa’s core values and what they can do for their members.
The new and improved web and Messenger chatbot
With the newest and latest version of the bot, the aim for Krifa is to include all members and non-members and to avoid any confusion as the content was made simpler and more manageable.
“We created a new version this year, a new version where we were available on Facebook and as a webchat, and we created it out of 8 different areas, so it was easier for the members to find the right answer.”
Now the new chatbot version is implemented at the bottom of Krifa’s homepage, and it opens full screen on the browser. In the near future, the website bot will be turned into a pop-up chat widget directly on the page. With BotXO you can have all your content in one centralized account and then use it in all the channels you want your bot(s) to live. That means for Krifa, that the old Messenger bot is updated with the same content as the website chatbot.
A key feature in the new version is that the members can easily pinpoint which topic in the bot is relevant for them, hence, the chatbot is now an improved bot that offers a nice and intuitive customer journey.
Using data to improve the customer journey
The way Krifa maintains and improves their chatbot is a great example for other companies that want to constantly improve their bots with new data. In order to build an intelligent chatbot that will give users a good service experience, companies should expect to allocate some time for improvements. It’s important to make sure that the bot is up to date with the newest information and fulfills the customers’ needs.
“Right now, there are 7 people, including me, that are working on the chatbot. And we are from 4 different departments: we have some from the Trade Union, some from the Unemployment Fund, one from Customer Service and one from the Digital Department.”
The different departments joined forces in order to gather all their knowledge to ‘feed’ the chatbot with important information from each area respectively. But first, they had to pin down exactly what questions the customers needed answers to:
“The new version had a focus on the data we collected: what do our members talk about in support and how can we use that data to create the new chatbot? So, we created it out of 8 new pillars with the most frequently asked questions.”
For the new version they collected data from their Customer Service department to figure out what questions their customers most frequently ask and they narrowed it down to eight different areas that the chatbot should cover; unemployment benefits, employment, how to register unemployment, job search, stress, how to submit a benefit card, membership and how to get in direct contact with an employee at Krifa. This way Krifa had managed to ‘unsilo’ their company by bringing the different business areas together in one bot, allowing members to get instant help regardless of which department their question belonged to.
NLU and intelligent chatbots
“With the old bot we realized that almost all our members start writing a question, but the old bot couldn’t take care of that. So, we created the new bot and started using NLU.”
NLU stands for Natural Language Understanding. It is an AI technique that BotXO has spent + 10.000 hours on developing. With NLU the chatbot can understand what humans are saying with their natural language. This approach is called a text-chat. This way the user doesn’t have to only follow the regular click-chat communication flow where the user will be guided through the chat using buttons with suggested replies. Instead, the user can write freely and ask a question. The bot provides an answer based on the input. The AI even takes into consideration that humans make spelling mistakes or sometimes have bad grammar. When Krifa’s members write a question to the chatbot, the chatbot understands the intent or aim of the question and can automatically provide the user with an answer.
“After we started using NLU, it was better all around, so when our members wrote to us, we could understand what they were writing about and we could give them the right answer and the chatbot also knew, if it couldn’t answer it, it had to send a message to our Customer Service.”
In order to create an intelligent chatbot solution with NLU, it is essential to use domains, intents, sentences or keywords that identify certain sentences, and to have an understanding of how the users express themselves in what is commonly called ´natural language´. This way the chatbot can cover all the themes that the users will want to ask about.
BotXO has developed pre-built domains and intents for companies to get started quickly and easily with NLU. It is also a tool for companies to build up on their own intents related to their own field. This is what Benjamin and his team focused a great deal on when building the newest bot.
“We built our own intents because some of the areas where our chatbot can deliver answers is specific for our kind of company. We have built for instance ‘happiness at work’, which is our mission, and ‘injuries at work’, so we can take care of people when they write to us.”
To whether Benjamin would recommend other companies using NLU, the answer is clear:
“If I would recommend using NLU? Yes, of course. Because for us it was the reason why we have a chatbot (…) and now it’s much better.”
Blending bot and human support
In some cases you could want to involve a human agent directly in a conversation with an end user. This is can be done live by transferring the chat to an available human agent, or by the bot automatically passing on relevant information so the human agent can take action when she is available. This is a way to seamlessly blend automated service with human service to ensure an effortless customer experience.
“The main reason why we need human handover is that some of the questions we get are not something that the bot can answer. The bot can’t answer how many days of dagpenge [unemployment benefits] you have left or when you will get the payment, or where in the process your case is (…) So, there are still some questions that the bot will not be able to answer but then we can provide the member a good service to find out where they can find that answer.”
When it comes to questions that are concerning a member’s specific case, the bot doesn’t have access to that level of personal information yet. Therefore, Krifa has provided the members with a human handover option.
Results and prospects
After implementing NLU in the newest bot, Krifa experienced a great number of members using the chatbot – another indication of the success of the Proof of Concept which started it all:
“We can see that in one and a half month we had 1.500 users and almost 3.000 visitors so we can see that our members use that chatbot and we look forward to see how it will affect the organization.”
Krifa has gone a long way in just over a year:
“When we started building the version that we launched last month, we started with a blank canvas. We started to see how we can use the data that we have from our Customer Service to create a better bot, a bot that can help our members. And after building and designing that, we went from answering 13 % of the questions to now answering almost 50 % after only one month in the air (…) Our goal for our chatbot is that it can give the right answer in 75 % of all conversations.”
On a final note
Now we know that Benjamin is a big fan of chatbots, but would he recommend other organizations to implement one in their business strategy?
“I would strongly recommend other companies to start using chatbots, because the ability to serve our members or customers at any time of the day, I think is very important for us. Not just to give answers during the opening hours but that we can give answers to all our members or to people who are not a member yet, and that we can give them the right answer in the middle of the night or even during Christmas Eve.”
Whether companies are small or big, with the implementation of bots there is a huge range of opportunities and benefits to follow.