Step 1: We design the bot
The first questions we ask are: What are our goals and use cases? In other words: How should the product owner’s business benefit from it? What will also make things easier for their customers? If we know that, we can draw the strategic framework. What are recurring topics and requests that cost company employees time and where a machine could help? Who is the target group of the chatbot? Which digital channel do I put it on – the website or WhatsApp? What tone of voice does my bot use with customers? We work out the answers in a multidisciplinary team with the product owner together in a bot canvas workshop. With this fast but clean basic framework, the dialogue design and exact content can be planned and implemented.
Especially in challenging times, a chatbot can pragmatically and efficiently handle an increased volume of requests on support channels.Carolina Cozacu, Senior XD Consultant, Ray Sono
Step 2: We teach him to speak
A question that is immediately asked is do I as the product owner have enough content for a chatbot so that I can go live quickly? The answer is usually: yes. It is a fallacy that chatbots must be perfect and must provide a detailed answer to each user’s question. Sometimes it’s enough to navigate the user to the right place or to give them the appropriate service channel so that they have a good experience with the bot, and in end effect, the brand.
Thus, the “critical number” of questions the bot can answer doesn’t have to be overwhelming. Once this is clarified, content – i.e. text modules in natural language, dialogue processes, and multimedia elements – can be produced as the core of the chatbot.
For a “starter chatbot” it helps if the most important texts and images are already available in digital form, so that they only need to be adapted, not created from scratch.
Step 3: We build it and go live
It is usually very clear which type of chatbot software makes the most sense for the specific case. In the meantime, various software providers sell chatbots that have different levels of complexity and licensing models. We would be happy to help you with your selection.
Simple systems are suitable for a fast starter bot, which even non-developers can set up intuitively.
To do this, you only need some technical know-how and the desire to tinker around. And then you can get started. You have to build dialogue trees, train and test language understanding, and finally integrate the chatbot into its planned environment using a simple code snippet. After a clean test, nothing stands in the way of the fast go-live.
Tips for afterwards
Using the direct connection to customers
Chatbots are always on the pulse of the customer and understand their every emotion. This can be risky, but it is above all an opportunity to identify really relevant issues without much preparation. Give your customers the opportunity to enter their questions, answer them initially only by referring them to an alternative contact option if the chatbot doesn't know the answer yet, but keep an eye on these valuable impulses and gradually add them to the bot’s knowledge base. In this way, the “starter chabot” will grow step by step and offer customer-centric communication with increasing added value.
Better today than tomorrow
Chatbots can be very simple or very complex. Important: They don’t get really good until they learn from real interactions with your customers. As with most digital products, continuous development and editorial support is quite useful. This is why it’s better to start now with a little bot and learn than wait too long.
We would like to accompany you on your chatbot journey. Together, we are laying the foundation for automated customer communication – with the “starter-chatbot”. If you are interested, we will always be at your side with advice and action!
Would you like to have some chatbot inspiration? It did not come into being in five days, but the DB bot Kai, who lives in the “Help & Contact” area on bahn.de and in the DB Navigator app, was implemented by us and is continually improving.