The popularity of chatbots reached its peak in 2017. With the hype surrounding of Siri, Alexa and other virtual assistants, the size of the chatbot market is expected to grow from $2.6 billion in 2019 to $9.4 billion in 2024 – at an annual growth rate of nearly 30%.
Mathieu Barthelemy, Lead Product Manager Digital Banking at equensWorldline, has been working in Digital Banking for more than 10 years. He noticed that chat- and voicebots were a hot topic at the events and summits he visited before the Covid-19 crisis. This may be due to the rapid and ongoing evolution of conversational AI, which can respond directly to new situations and answer clearly in everyday language.
Many organisations are using the crisis to explore ways of increasing resilience in the future, such as digital customer contacts. Because bots fit so well in this kind of strategy, here Mathieu is offering three lessons to keep in mind when developing a chatbot.
1. Text has the priority for the time being
The first step in developing a chatbot is formulating a clear goal and matching tasks. What exactly do you want to use the bot for? What tone of voice should a bot use with your customers? During this early stage, you should also determine whether you want to use a text-based virtual assistant that is included in your web/mobile app or available on social networking platforms such as Messenger, WhatsApp, or Telegram. Or, perhaps, a voice-based virtual assistant, such as Alexa or Siri.
The first advice is to start with a chatbot assistant for your text channels. There are two important reasons for this. First, end users currently expect services via bots to run mainly through a text channel. As a company, you need to be where your users are. With that in mind, take a look at WhatsApp for example: according to statistics, WhatsApp users sent 65 billion messages per day.
The vocal channel is the channel of tomorrow and will grow in the next 3 years but is not there yet. That is why it is better to focus on text now and build up expertise in this area. That will make the transition to voice a lot easier afterwards. Second, it is still difficult to manage reliable biometric authentication of end users with a voice channel. If you have a choice, choose text.
Keep in mind that the main difference between text and voice is the messaging: text messages can be viewed over and over again, which makes it possible to keep hold of the context for a longer period of time. With voice this is not possible, so the messages must be shorter and much more to the point.
2. Building a good conversation UX is different from building web/mobile apps
Setting up user-friendly web and mobile apps can both be quite a challenge. There is, however, also a big difference between these two interfaces. With a chatbot, you cannot use a lot of visual or graphical components compared to mobile apps. This means that developers need to go a step further when building chatbot channels to provide customers with a great conversation experience.
One of the problems is the fact that conversation becomes a funnel with too many steps and a lack of connectivity between them. That’s why several companies use a design approach with which they can take up the challenges below:
- Avoid long conversations.
- Bring more clarity into dialogues/conversations (simplicity is key).
- Make the design of new conversations smoother and faster. For this, it is important to consult people with specific skills, such as marketing-, IT- or UX-design skills. All this with one goal; to define a smoother conversation with chatbots.
3. The bot’s future development is a strategy priority
An ‘exploration strategy’ is essential when launching a bot. Working on the accuracy (training) of your bot should be a much higher priority than wanting to rollout your bot on other channels at the same time. Your biggest challenges lie in encouraging end-users to engage with the bot and improving your bot on a frequent basis. This enriches the bot’s ability to understand end-users and avoid frustration.
People think that improving a bot is an automatic process, but you have to train the model. And this training part is really important. If you develop a text-based bot with an exploration strategy and a careful examination of the user experience, your investment will lead to long-term customer engagement. But how do you train the model?
Build custom conversations
Fortunately, there are solutions in the market that help speed up the training process. Formulations, sentences, answers already trained and integrated in these solutions to fasten the time to market with only personalisation workshops with clients. In addition to these pre-trained conversations, it is also possible to use their conversational expertise to define and build custom conversations and address specific use cases.
Deploying chat- and voice-bots induces many interrogations and challenges, notably related to the proper understanding of the user’s query as well as the choice of the most appropriate answer. But it also enables banks and other financial players to securely improve customer journeys with tailored and contextual answers while significantly reducing customer support costs. Far for replacing human interactions, conversational interfaces are relieving pressure from advisors and help them focus on more complex and valuable services. For this reason, chat- and voicebots are slowly being embraced in the finance world.
