To meet customer demands and strengthen customer experiences, Schibsted introduced a chatbot.

The customer center’s challenge is complex. Traditionally, the customer center has been a cost center, which has sought to resolve the requests as quickly and efficiently as possible at the lowest possible price. This is occasionally at the expense of the quality of work carried out. But since customer loyalty has traditionally been strong, it has been possible to prioritise downward customer service as a function.

In recent years, customer service expectations have risen significantly, and the customer compares the service he or she gets with what is experienced by best practices, regardless of industry. The customer has high expectations of being informed of what is affecting the customer relationship and can potentially contact the customer center at any time throughout the day. New technology and smartphone access wipe out the time limits, and the customer expects to respond to a chat or e-mail regardless of the time.

The share of self-employed inquiries has risen steadily, but the cost base represented by the customer center is still significant for the company. To ensure profitable finances, it attempts to reduce this share.

Prior to the development of the chatbot, three main challenges were identified in the company:

1. How can we solve the customer’s challenges more effectively and be more proactive?
2. How can we extend opening hours and be more accessible to the customer without increasing the cost of the business?
3. How can we increase customer experiences and strengthen customer relationships with Schibsted across the brands?

What opportunities did we have?

The company believed a robot connected to the chat solution would be a great opportunity to meet the challenges.

A chatbot could provide a better customer experience by giving the customer immediate answers and questions, as well as ensuring increased availability, because a bot does not depend on crew in the back. If the robot was optimised with the right content, the company could offer a 24-hour customer centre.

Finally, a robot solution could also be proactive for the customer and provide information the customer did not necessarily search for himself.

Advantages of implementing a chatbot

The hypothesis is the chatbot will remove the traffic jams and this will give the customer centre the capacity to raise competence in other areas.

New roles and other skills

As a result of the implementation, we defined three priority areas to increase the customer service center’s quality.

1. Increased advanced basic skills: The chatbot removes repetitive inquiries and leads to some sort of expulsion of the more skill-requiring tasks. This requires that the level of competence must be at the level seen by experienced consultants.
2. Better writing skills and technical skills: Optimising the chatbot requires high textual understanding and technical competence. This is the competence we must strengthen.
3. Analysis expertise: It would be interesting to analyse the customer’s interaction with the chatbot. This requires analytical competence and understanding.

A lower need for seasonal work also means the constant need for new worker training will decrease drastically. In addition, the customer consultants who are working will have the opportunity to increase their skills, which in turn will provide a better customer experience and strengthening of the brand.

The potential risks

• Potential for impaired brand: If the customer finds the chatbot answers incorrectly or is not human-like, this may impair the brand. We, therefore, need to ensure a quality information flow to the editorial staff, as well as have a strong focus on optimising content in the chatbot.
• Failure to scale down the number of employees: Consultants’ time may be used for something else or they may have reduced efficiency due to the decrease of pressure on them. Therefore, a reduction in the number of employees should follow the volume increases in the chatbot so we maintain a proper, effective level.

Based on the definition of the potential advantages and risks of implementing a chatbot, the technical solution was defined.

Technical solution

The company wanted an advanced learning robot that did not fail in the response to the customer when asked difficult questions. The chosen solution is based on historical chat dialogues and FAQs and uses new dialogues as training data. This generates ongoing better responses.

To ensure an optimal customer experience, it has the functionality of a human handover if the tasks are too complicated for the chatbot. Finally, we wanted a robot built on the Norwegian language model, not based on translation, for better accuracy.

Results

Before implementing the chatbot, we saw a clear trend of increasing inquiries that would need to be manually processed. This trend has turned, and we see a reduction in this type of requirement since implementing the chatbot.

In 2017, there were 70,369 chats. In 2018, there have been 70,939 chats; the total number of chats that includes the chatbot is approximately 85,000.

As the number shows, total traffic for the year is similar to last year, but we also have a significant volume processed in the chatbot (approximately 30,000 calls from start-up). Around 50% of the traffic is already a part of the 70,000 treated this year, but the remaining 15,000 is not. These are handled by the chatbot. This amounts to approximately 18% of the total chat traffic.

In addition, in 2017, we received 113,428 customer e-mails. In 2018, we have received 121,833. Total e-mail volume has increased throughout the year, but lately the curve has had a negative trend. That’s a good thing! This may mean we are seeing effects of channel changes from e-mail to chatbot.

E-mail accounts for 20% of total traffic. By comparison, chat accounts for approximately 8% of total traffic. The greater the proportions of chat and chatbot, the more efficient we are at handling the inquiries. E-mail is a time-consuming channel.

Priorities

To achieve the effects of implementation, the company needs prioritise the following actions:

1. Optimise content of the chatbot.
2. Analyse chat dialogue to identify the need for new functionality in the chatbot.
3. Develop functionality in the chatbot.

Additionally, a larger part the requests must be turned from regular e-mail to digital surfaces where the chatbot is profiled, such as on the content pages. In that way, the inquiries to the customer center will gradually be resolved by Artificial Intelligence and chatbot.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .