Knowledgebase-Intents

Find out more about the new knowledgebase intents, how they are use, in which cases it is advisable or not advisable to use them and what part generative AI plays in it.

Knowledgebase intents in general: 

Knowledgebase intents are an alternative to traditional editorial intents. This new form of intent makes it possible to provide individualized and dynamic answers. The process is based on generative AI or a large language model. 

In contrast to the answers provided in the editorial intents, the answers in the knowledgebase intents are not generic and repetitive, but react dynamically and individually to a user enquiry. 
While the answer of an editorial intent is based on the previously created response content, selected sources form the knowledge database for the generative AI: 

It is currently possible to store websites and PDF files as sources. It should be noted that editorial control over the sources always exists, but not over the specific output in the chat. 

Knowledgebase intents are activated via the 3-dot menu in the response editor. Click on the "Knowledgebase intents" button to open the option to change the intent from editorial to knowledgebase.

All previous contents of the answer will be lost if you change the intent-type! It is therefore essential to save the content of the editorial intent. This is the only way to restore the content via the change log (see here).

After the changeover, the editor of the knowledgebase intent opens. Click on the "Add" button on the right-hand side to open the input mask. Adding is done by entering the link (if it is a URL) or by upload/drag-and-drop (for a PDF file). 

The added source(s) appear in the source list. If there are several sources, it is possible to filter by source type. The filter is set by clicking on the PDF or website button. The sources in the topic list form the basis for the answers then played out in the live chatbot.

As a result, this means that only the information from the stored sources is included in the response. 
If a file is no longer up-to-date, it must be replaced in order to prevent outdated information from being displayed. 

Playground

You can test the AI in the Playground. Questions entered here are processed by the AI and it plays out answers that would also be played out in the chatbot. The AI also indicates here which questions it does not know the answer to or for which no answer can be generated from the inserted sources.
Answers can still be rated positively. The positive rating encourages the AI so that this answer will be displayed in future for similar/identical enquiries.

Knowledgebase topics: Best and worst use cases

Knowledgebase intents are primarily used to reduce the editorial effort involved in creating answers. They are therefore generally suitable in areas where high editorial effort meets low query volume (so-called long-tail queries). 
For intents with a higher volume of enquiries, an evaluation of the response documentation is necessary beforehand. If the documentation is good, the knowledgebase intents are also suitable here. 

Knowledgebase intents are also suitable where the answers are very repetitive and good documentation exists at the same time.

However, knowledgebase intents are unsuitable where the answers are dynamic and based on complex documentation, e.g. questions about individual product KPIs. 
Knowledgebase intents are also unsuitable if the answer contains sensitive information, e.g. dosages of dietary supplements or information on allergens. 
Finally, a knowledgebase intent is not recommended if only a small amount of editorial work is required to create an answer.