Chatbot conference 2019 (Utrecht)

HW: Maarten, you attended the chatbot conference last year. What did you think of this edition?

MV: I have to say that last year I felt more positive about the conference than this year. At the time, everyone was very enthusiastic about the possibilities of chatbots. A few more years and customer services could be closed. Cost savings would benefit the companies. This year, however, I noticed a more conservative mood.

HW: After hearing your enthusiastic stories from last year, I went to the conference this year. But to be honest, I found the presentations rather disappointing. The issues I know of with chatbots were covered very superficially, with no real solutions. Perhaps that is also the role of this conference: mainly to inspire and not to inform too much.

MV: Sure, but a tip like 'Fill your chatbot to start with the FAQs' is not very inspiring and it is of course also questionable whether you actually help customers with preconceived questions.

HW: Exactly! The minimum you need to know is what customers really ask and want to know, in order to determine their intentions. You do that through research or by collecting real customer data and not by having a marketing team come up with it. In any case, you will not get a well-functioning chatbot with this.

MV: That last aspect is interesting. I now clearly noticed (as a.s.r. pointed out in their presentation last year) that a chatbot where customers can ask questions is not going to be it. Things are clearly heading in the direction of a decision tree. For example, the customer is presented with five topics and for more difficult questions, the customer is connected with an employee. I also saw this during the presentation of one of the suppliers who claimed great success with their chatbots. Customers could ask only three questions.

HW: Yes, I began to realize that chatbots will no longer replace the entire customer service, but will provide an extra communication channel with which you can serve customers in a different way. The business case for a chatbot is not based on call reduction (and therefore cost reduction), but on a better customer experience. And this does not only apply to customer service, but also to commercial applications. My experience with the CZ chatbot also showed that KPIs solely focused on call reduction are simply unfeasible. Managing on quality is also a great challenge. During the second keynote about the new chatbot Nina from Nuon, a challenging target score of 9 was presented. Knowing that the average feedback from chatbot users, even on correct but unwanted answers, is significantly lower in practice, such a score as a goal seems the perfect recipe for disappointment.

MV: : Yes, that's right. But if you have employees in the background who can help with difficult questions (or at least questions that the chatbot does not know how to deal with), it does contribute to a greater customer satisfaction.

HW: There were several suppliers at the conference who offer such a (warm) link with a 'live chat', so that's hopeful. For organizations, of course, this implies more than just technology. You also need to have enough "agents" with the right skills who are able to take over conversations. I wrote an article in 2017 about chatbots that require artificial intelligence (AI). What have you actually heard about AI for chatbots this year?

MV: Just like last year, AI is the magic word this year. Everyone's talking about it, but if you keep asking, "How does it work?" or "How do you deploy it?" you run into a wall of silence. It's like asking a cook for his recipe.

HW: Yes, there is change in the air of the chatbot market, but several speakers – also during break-out sessions – don't really come up with details that will get you further. They are cooks who don't give you the recipe but tell you to just come and eat. By the way, I thought that was the tendency of this conference: suppliers who give you just a taste to get you interested in doing business with them.

MV: Is that bad?

HW: No, that's okay. Particularly when you don't mind attending lectures without real answers, but only stating the obvious. It may be informative if you're just getting started with chatbots, but not very sustainable in the longer term.

MV: So the chatbot hype is past its prime. What's left in the valley?

HW: As far as I'm concerned, the search continues. Towards good applications of chatbots, to working technology and to ways to get and keep a chatbot up and running. But certainly also towards an approach to set up your content in such a way that not only a chatbot can use it, but also the other communication channels of your organization. Because I think that's where the real solution lies: a chatbot should not be a separate channel, but part of a coherent omnichannel content architecture.

Co-author: Maarten Vos