Why is AI so important for a chatbot?
Die KI gives the chatbot a form of intelligence. Because it is only thanks to AI that a chatbot is able to understand complex messages, assign them to the right topic and thus deliver the correct answer. In addition, a chatbot only learns and develops independently if it is AI-based. Of course, there are many Features and benefits that arise from AI. One of these features is dreaming.
What is dreaming?
Chatbots are often criticized for understanding little apart from trained knowledge and are therefore unable to process or even reflect on unknown content. This is different with moinAI. Thanks to Dreamings is moinAI also able to process content or user requests that go beyond the knowledge learned. In short: moinAI tries to make sense of everything, including user inquiries that are still unknown to AI. So when topics and phrases arise again and again that artificial intelligence is actually not yet familiar with, it can reflect this and cluster the unknown formulations on topics. These topics are suggested accordingly to the chatbot owner, i.e. the company that uses the chatbot, so that the chatbot is trained with the new content and can now also address these topics.
practical examples
Admittedly, the dreaming may sound a bit abstract for now, so it makes sense to use the feature based on two scenarios to explain.
- Expanding topics at a window manufacturer
A company that manufactures windows integrates the moinAI AI chatbot solution into customer communication. The AI chatbot is now being trained on various topics, for example in the case of the window manufacturer: “Get an offer”, “warranty”, “costs” and “window sizes.” These topics were selected by the company because customers and interested parties are particularly frequently asked about them.
When selecting chatbot topics, the focus is initially exclusively on the most important topics, so that the chatbot is not trained in advance with the entirety of potentially relevant topics. In the case of the window manufacturer, for example, the question of the window code was not initially included. However, the chatbot now regularly receives user inquiries such as: “Where is the window code?” , “Where is the window number?” , “I need the window type number”, “Where can I find the model number?” - and did not provide an answer to this question.
However, the AI registers that although it does not know this topic and cannot answer it, it is relevant for users. So she clusters the said queries on a topic and suggests this to the window manufacturer's chatbot owner. As a result, the company learns that the topic of “window code” is important for its customers. As a result, it can use these insights and insert the appropriate answer to the topic of window code into the AI chatbot so that it can now also help users on the topic in question.
- Extension from a customer service case to a marketing & sales case at an insurance company
A second scenario could be exemplified by a insurance be shown. The insurance company has integrated the chatbot into customer service, thus reducing the support volume. The chatbot was trained on topics such as “report damage”, “login problems”, “change names”, etc. These are exclusively service-related topics intended to help existing customers.
After the AI chatbot has been live for a few weeks, it registers that questions are being asked more and more by potential customers who are interested in insurance. Although the chatbot is also used by existing customers, inquiries repeatedly arise that are more likely to Sales & Marketing should be assigned and not to customer service. Examples include inquiries such as “I need advice on an insurance plan” or “Why should I choose the premium insurance package, or does the basic rate make more sense? ”. The chatbot did not provide any answers to these questions, but the AI registers that these queries appear to be relevant, otherwise they would not be made.
As a result, the AI clusters these inquiries on topics and suggests them to the insurance company. In this case, the AI chatbot would therefore point out to the company that the chatbot is obviously not only used by existing customers, but potential customers also expect product advice from the chatbot.
Insurance companies can now expand the chatbot to include the topic of “product advice,” including “picking up” potential customers, informing, inspiring them and ultimately increasing the conversion rate and increasing the number of leads. Especially in this case, moinAI has another advantage: So-called bot forms can be used to generate leads automatically. If a chatbot user is interested in a tariff, the chatbot can automatically request their contact details and forward them to the insurance company's CRM system so that sales staff can continue working with the generated lead.
- Improved customer understanding in publishing
In addition to a classic sales company or insurance company, illustrating the generation of customer insights using a publishing house is also very helpful. Let's take a closer look at the AI chatbot from the publishing house Spektrum der Wissenschaft, which is actively used in practice to generate and process customer insights.
Spektrum's chatbot is used on the landing page via subscriptions and was initially addressed to classic questions, such as “What's included in the subscription?” or “When will the next issue be out?”, trained. Some time after the Chatbot implementation There were other user questions that the chatbot had not yet been trained on, such as “How many pages does an edition have?”. The reason for this question is that people who are not yet familiar with the publisher's products want to know how much content they get for their money. Even though the chatbot was unable to give a clear answer to the question at the time, it noticed that the topic was relevant for users because it was recurring and recurring. The AI then suggested to the chatbot owner (in this case spectrum) that the new topic be included in the chatbot as a question.
In addition, Spektrum der Wissenschaft has also taken appropriate measures on the website: For example, an info box has been included on the table of contents pages of all publications, which addresses the basic data of the publications: What can users expect when publishing? How many pages does the publication consist of? How often does the publication appear? This is how the chatbot now ensures that the question “How many pages does an edition have?” Users hardly ever care about it anymore because the answer to the question can already be found clearly on the publisher's website.
“How can I change my subscription? ” (e.g. from a print subscription to a digital subscription) is another example of a question that also came up among visitors to the Spektrum website following the chatbot implementation. In this case, too, the topic of “subscription change” was not only included in the chatbot, but was also integrated into the contact form for the support team on the website so that they can answer appropriate inquiries more quickly and, if necessary, with fewer queries.
This is what dreaming looks like in the hub
The topics clustered by AI are collected and processed in the form of dreaming suggestions in the hub, i.e. in the backend of the chatbot. There, chatbot owners can view the dreaming suggestion, enriched with typical user inquiries, i.e. various variants with which the topic is queried.
It is also clear how relevant the topic is, i.e. how often this topic is asked about. Since this is a dreaming suggestion, the topic is only live when the chatbot owner prompts it. If the topic is not yet live, it is worthwhile to redirect to a topic that is already live and has a similar intention.
The following image shows that events seem to be requested frequently. This is registered by the AI and created as a dreaming suggestion. However, since the topic isn't live yet, it makes sense to redirect to a similar topic, calendar, until then. This could answer a few questions for now until the events topic is online.
More insights and knowledge about what really moves users and target groups
Dreaming from moinAI has one clear advantage in particular: It tells the company more about what users really want and expect. By not only collecting topics that can currently not yet be processed by a chatbot, but also intelligently clustering, companies learn exciting things Insights about your users.
Using “dreaming”, the chatbot's AI is able to identify through analysis and self-reflection what users are also interested in and which topics the chatbot should be expanded on in order to further improve the user experience. The AI chatbot thus gets to know the target group or website visitors in the best possible way, opens up new topics step by step and adapts to the wishes and needs of users at the same time.
As explained in the previous example about the window manufacturer, the online sales organization Velux OSO, which is known primarily for its skylights, also uses an AI chatbot for B2C online sales. In this way, OSO manages to reduce live chat volumes and serve multiple markets at the same time. Klick hereto find out more about the use case.
