URL has been copied successfully!

Everything is love at the Conversational Conference

by Erik Bouwer

Everything is love at the Conversational Conference

by Erik Bouwer

by Erik Bouwer
URL has been copied successfully!

Love played an unexpected toll on the bol.com and Efteling sessions. The bol.com chatbot was presented as the love broker of the Netherlands. Maybe he can do something for Pennenveer, who is asked for his hand in marriage so often by guests at Efteling that a trap has been built for them.
Part 1 of an impression of a well-attended Conversational Conference by Frankwatching.

Text and image: Karina Meerman

conversational

Mark van der Heijden. bol.com

Thanks to customer service product manager Mark van der Heijden we now think fondly of bol.com. Their chatbot Billie is the largest love broker in the Netherlands, “because he asks the right question to the right person”. Bol.com's online store has grown in twenty years into a 'shopping platform' with 50.000 sellers and therefore also with 50.000 customer service departments. Some salespeople serve their own customers from the attic or with an entire department. Others have outsourced customer service to professional parties. Sellers range from sole proprietorships to multinationals. Then there is bol.com's own customer service, which employs 1.600 people, which increases to 2.500 during peak times.

Top 5 questions

Customers who have questions for bol.com look for answers on the self-service page of the website. They also find Billie there. Calling is not an option and will not be. Van der Heijden: “We want the platform to be the first entrance.” When he asked the audience which question was asked most often, they shouted back: “where is my package?” And that was true. Closely followed by 'I want to return something', 'I want to have something repaired', 'there is a problem with payment', and 'there is a delivery problem'.

Billie's choices

Customers who order on bol.com do not always buy from bol.com. Yet their questions must be answered correctly and that's where Billie makes his entrance. Billie works based on generative AI and does “hundreds of days of work in one day”. He determines who is the best employee to answer a question. To do this, he opens all data registers known to him. When asked: 'where is my package?' Billie checks whether it is a product from bol.com. If so, the track and trace code is known and the question can be submitted to our own customer service. Is the product from another seller, but sent via bol.com? To your own customer service. Is it from a seller who arranges deliveries himself? The seller must then answer the question himself in the seller portal.

Happy match

With 'intent recognizer' (what do they want), 'dialogue engine' (what do they say) and 'back-end APIs' (what do we know), Billie then matches the question to the employee with the right skills. Including all known data about the customer, the language, the product and the question/intent. A 'happy match' is the number of times a customer has a point of contact with one of the people at the seller or at bol.com. Van der Heijden: “Ideally that is 1. Then the question is answered.”

Control over quality

In the last two years, bol.com took IT back from outsourcing. “To be able to innovate on the platform and especially on that last part, the match.” Billie, on the other hand, has never been outsourced: “It was always ours.”

There is a seller program to inspire sellers to provide good customer service. The first step is convincing, with online training and providing NPS feedback. The second is seduction with the promise of the buy button. This is the holy grail of the platform: the yellow button with the shopping cart next to a product. The Buy Button winner has the best combination of price, delivery time and customer review score and the best chance of scoring a sale. Step three is to honor the agreements: if the scores are poor, a seller will ultimately be removed from the platform.

True love

Finally, the happily married Van der Heijden shared on the conversation conference “the bol.com recipe for true love”.

  • Turn your chatbot into a matchmaker;
  • Make sure that the chatbot only asks what you cannot yet know (link with your back-end);
  • Use all that information to make the right match;
  • Set up the back-end in such a way that the question ultimately comes to the people who can best solve it;
  • Make right matches a customer service KPI;
  • Give employees the tools to make an indelible impression.

What the future of generative AI will bring for bol.com is still uncertain. “We don't know whether the generated content is correct,” Van der Heijden said. “We are not currently using it for customers, but we do share the results with our employees. All AI-generated content is checked until the trust in it is high enough.” And Billie? In the future, it will increasingly become part of the entire customer journey and be present wherever the employees work.

The Efteling: wedding trap for entertainment

As long as the pen quill continues to write
Will Efteling remain open forever?
Here every human child is rejoiced
A memory for eternity...

Pennenveer is a monument on the territory of Efteling and the name of the chatbot that has been live since 2020. The fairytale forest that opened in 1952 with 10 fairytales and 50.000 visitors is now called the World of Efteling and receives 5 million guests per year. The vision for 2030: an international park where everyone experiences a 9+. This also has consequences for the development of Pennenveer, as became clear during the presentation at the Conversational Conference.

Functionality

Daphne de Klerk (IT&Digital program manager) and Every Smolders (product owner conversational & customer view) talked during the Conversational Conference about the growth of Pennenveer, whose personality is described in a style guide. He is enthusiastic, funny and eager to learn and wants to make guests happy. Penfeather – 'just call me Feather' – loves language jokes and fairytale references. Multiple 'conversational writers' provide the chatbot with text and content. De Klerk: “We are committed to personalization and do our best with the legacy systems.” The application of generative AI is now running as a pilot for customer contact employees to create conversation summaries. From the end of this month, people who are interested in working at Efteling will receive their own Pencil, containing the most frequently asked questions about working at Efteling. Visitors can, among other things, reserve restaurants in Pennenveer, request a wheelchair and view the times of shows.

Multilingual

conversationalThanks to Seamly and translation machine Deepl In addition to Dutch, Pennenveer also 'speaks' English, German, French, Italian, Spanish and Mandarin. Only five percent of all customer interactions with the Efteling Contact Center (ECC) were in a foreign language, but a translation option matched the international vision and the ambition for a 9+ experience for everyone. International customers provided new questions and insights for the former fairytale forest, such as 'what is the nearest airport'.

Translations are always read back. When it turned out that Veer automatically translated the Bird Skirt attraction as Bird Skirt, an exception list was quickly created for the translation machine Deepl. All attractions retain their Dutch names. Answers with humor and puns that are difficult or impossible to translate will receive an alternative version.

Hotel industry

Pennenveer has recently also become available for people who spend the night in Efteling. The hotel industry came with its own challenges. The questions are more detailed: what are the dimensions of the beds, is there a wok, can I have groceries delivered? Pennenveer also suddenly had to deal with reports. From 'I don't have a towel package' to 'I tripped and now the stair gate is broken'. Smolders: “We became a facility channel and we were initially not prepared for that. Before the park, the ECC had data on which we could write content, but not here.”

Wedding trap

The character Penveer turned out to be an unexpected source of data pollution. For example, visitors surprisingly often asked if he wanted to marry them. There was no standard answer to this and the colleagues on the live chat became a bit faint from the funny interactions. “We analyzed it: after 'will you marry me' there was never a serious question again,” says De Klerk. When visitors ask a question that is recognized as entertainment, they are now led to the virtual play corner. There they can enjoy videos, quizzes, jokes and stories. Content is added every two weeks, because some guests can't get enough of this digital entertainment. “Internally the play corner is known as the trap. Sometimes people can't figure it out anymore.” (Ziptone/Karina Meerman)

Also read part 2 of the impression of the Conversational Conference about Rabobank and GEM.

Interesting article? Share it!
RSS
Follow by Email
Whatsapp
LinkedIn
Copy link

Also interesting

Case, Featured, Technology
Top