Messaging apps are extremely popular and are installed on billions of consumer smartphones. On the other hand, more and more companies are using chatbots and voicebots. For customer contact, both developments are expected to reinforce and complement each other. What's up with that? In the run-up to the CCMA Digital Transformation event on June 15, Ziptone will discuss this with Michiel Gaasterland (CM.com) and Sander Hesselink (Telecats).

 

A while ago I asked the chatbot of an energy company whether this company also supplied a charge card. The chatbot came up with the response: 'Unfortunately I don't understand. Please try again or ask another question.' The latter is of course hilarious: as if you are helping the customer with this.

Michiel Gaasterland has to laugh. “Yes, in this story at least one person is innocent. And that's the chatbot. The differences between companies in how they deal with chatbot technology are large. It depends on the business model, the culture, the maturity, the ambitions. Sometimes price is the dominant element in a company's proposition. I myself chose an energy company where it can take eight days before I receive an answer to an e-mail, but I have factored that in.”

Customer contact in relation to customer lifetime value

Michael Gaasterland is an 'evangelist' at CM.com, the fast grower and specialist in the field of 'conversational' and listed on the stock exchange since 2020. The company sends 8 billion messages annually on behalf of its customers. “The customer lifetime value is a determining factor when dealing with chatbot technology,” Gaasterland explains. “What are the purchase frequency and average transaction value within the organization? In fashion, for example, the average order value is 70 euros and the purchase frequency is on average 4 times a year. Every customer interaction is then an opportunity to further strengthen the relationship. But if you sell 20 euros worth of school supplies once a year, that becomes a lot more difficult and you need a very tightly organized e-commerce process. Customer contact then puts a relatively large pressure on your margin.”

 

Especially when using chatbots can be seen that things often go wrong. Why is that: is the knowledge base too limited? Was the approach of that implementation chosen incorrectly? Is the chatbot presented as omniscient when it is not?

Gaasterland: “I'm afraid all the points you mention are relevant. How successful companies are in deploying new technology also has to do with the maturity of the organization. Do young companies have the right skills and the right experience? Or are they doing business with the wrong partner, who primarily sells technology and doesn't really help with implementation?"

Sander Hesselink (Telecats) adds: “And to what extent has a company set up a feedback loop, so that chatbots learn from questions that are not properly handled by a digital system? And another factor: how long has customer service been using a system? So how mature is the chatbot itself? Its age is of course not visible to the customer.”

 

The chatbot is good at a few things, but the customer doesn't know what those are. How do you see this?

Gaasterland: “You can choose to use a chatbot that can only do five things, for example, but with which you can eliminate 30 percent of your customer contact volume. I would say: then manage the expectation well. For example, by having the chatbot say 'I just left school and can't do that much yet', or by explicitly stating that the chatbot specializes in a limited number of tasks. You should also make the development of your chatbot part of your roadmap: into which phases do we divide this project and where do we want to end up with our chatbot?”

 

So you should actually treat the use of a chatbot as an evolution, a step-by-step process?

“Relatively inexperienced chatbots indeed depend on their training. But some things you can immediately put in as a standard skill,” says Gaasterland. “For example, requesting customer data, so that the employee who is going to help you already has the necessary information.”

“You have to train a chatbot application over a longer period of time, so that all flavors and topics are covered.”

Hesselink: “Another factor in this is that in some branches different types of questions are received in the summer or winter than in other parts of the year. So you have to train a chatbot application over a longer period of time, so that all flavors and topics are covered.”

Gaasterland: "Something that is immediately possible is to indicate that you can ask questions to the bot, but that they will be answered later by e-mail - for example because customer service is closed."

Hesselink: “There are all kinds of options for deploying and facilitating such a channel switch. For example, you can provide a call-me-now request with a link that provides the customer service representative who is supposed to call the customer with all relevant information. In a similar way, with the VGZ voicebot, a number of fully developed customer journeys are placed behind the voicebot. For example, a question about the reimbursement of glasses can be handled fully digitally and automatically.”

 

What about the classic dichotomy of rules-based and AI-based chatbot? Is that distinction still relevant?

“Basically, chatbots work according to a model: now me, now you, now me again, now you again,” explains Gaasterland. “The big difference between rules-based and AI-based is that the latter does its best to recognize a customer's intent. As soon as the question has been properly mapped out, an automated answer rolls out. Generative AI such as ChatGPT will ensure that asking questions in return becomes more extensive and better. Challenges aside, such as how to handle unlocking, excluding and protecting data, ChatGPT is one of the most exciting developments in customer engagement I've seen in my life.”

Rules-based chatbots are just as rigid as an IVR

“Scripted or rules-based chatbots have a very limited role in customer contact. It is suitable for triage-like tasks, can play a role in routing and can work with FAQs. In practice, this means: carrying out a number of specific, clearly defined tasks. When the customer leaves that defined path, the rules-based chatbot crashes. With an AI-based chatbot, the customer has many more options to direct the conversation itself.”

To illustrate this, Hesselink makes a comparison between IVR and speech technology. “The IVR forces you to choose a well-defined option, voice routing, based on your spoken question, looks for a follow-up action that is expected to best match your intent.”

“The impact of new technology is overestimated in the beginning and underestimated in the long run.”

Both agree that the chance that consumers will ignore the rules-based chatbot is increasing as the AI-powered chatbot continues to develop. On the other hand, according to Gaasterland, we must take into account the so-called 'Law of Amara': the impact of new technology is overestimated in the beginning and underestimated in the long term.

Gaasterland: “There is still a lack of insight into exactly how ChatGPT works, it is a black box. Another major objection is that it is currently fed with data that we have no insight into. You can also see it as one of the largest data mining operations ever.”

 

There are many complaints about chatbots, but Ziptone readers will also want to know about the successful examples.

Gaasterland: “In so-called 'emerging markets' you will find large groups of consumers who have skipped the intermediate step of e-mail: they have gone online via smartphone. In the Middle East, for example, financial institutions in particular are well advanced with the use of chatbots. Vodafone has an excellently equipped chatbot in the Netherlands. At companies where chatbot technology works well, I already hear that they speak very little to the customer. They then start thinking about a broader engagement strategy.”

 

As chatbot technology matures, should you think more about customer engagement?

Gaasterland: “Indeed, then you have to think about coordination with other channels, for example. More than five billion people now have a messaging app on their phone. This has major consequences for customer contact. It also creates opportunities: when you proactively inform your customers about an offer, you are immediately engaged in a dialogue. This means that the boundaries between sales and marketing and service are blurring. Big Tech is investing heavily in the messaging channel: think of Apple Messages for Business, Google Business Messages or Meta that wants to add more functionality to WhatsApp. Instagram has its own messenger service and there will be its own payment solution for in-channel payments. This means that there are more and more possibilities to organize your entire customer journey within conversational channels. ChatGPT shows how easy it becomes to ask questions. 'Do you want to buy a mattress? What kind of mattress? I'm putting a few side by side here. Do you want to take a closer look at that left one?' That's where it's going. Incidentally, there are European regulations that require 'message interoperability' to be arranged within two years, so that I can send a WhatsApp message via Signal. Conversational will become an important interface between companies and customers.”

 

Text-based communication lends itself well to real-time translation and automation. A chatbot can therefore also be located on the other side of that messaging channel.

Hesselink: “Companies will also have to invest in human capacity, such as the 'fallback' that is necessary for digital customer contact.”

Gaasterland agrees: “I don't see many applications in the near future in which no human being is involved anymore. There must always be an escalation option.” (Ziptone/Erik Bouwer)

Featured, Knowledge base, Technology concentric, Concentrix Catalyst, Telecats