Each month Knowledge Studio will give you AI recommendations that help you identify what to write about next. We will recommend topics that are based on recent issues your employees are facing. This is accomplished by analyzing the last 30 days of IT tickets and determining if there are any knowledge gaps and knowledge updates you can address.
Within the product you’ll see two types of recommendations:
- Knowledge Gap: Our “Knowledge Gap” recommendation will create a brand new article for you about issues your employees are submitting tickets about, which are currently not addressed in your knowledge base.
- Knowledge Update: When we detect that employees are filing tickets about a topic that already exists within your knowledge base, we’ll tag as a “Knowledge Update” for you and generate an article that addresses the issue that was reported. For the classification on how we’re determining a “Knowledge Gap” vs “Knowledge Update” recommendation we are looking at your employee facing knowledge articles.
Every month we analyze the last 30 days of IT tickets that were filed within your organization to understand what issues your employees are facing. From a technical perspective, our process for how we generated these knowledge opportunities comes down to four steps:
We use GPT-3.5 to create a summary for each IT ticket by analyzing the description fields within each ticket.
Summary example: “Employee has access to postman but no workspaces assigned”
Each ticket summary is then embedded using a fine-tuned MPNet model to group related tickets.
Each cluster in this figure contains a set of relevant tickets.
For each ticket within a cluster we use GPT to extract ticket notes to learn the actions taken to resolve the issue. We extract information across the following ticket fields: closed notes, work notes, and activity log. This process produces a summarized list of bullets that describe each action agents took.
- “Confirmed purpose of access and type of access needed”
- “Provisioned ‘Postman Workspace - Deployment’ access to Marcus Smith”
Given the ticket notes we extracted, per cluster we select the top 10 tickets that will be used for generation. Then, we task GPT to generate a knowledge article by leveraging all the ticket notes we extracted.
This produces an article that is ready for review to ensure the information generated conforms to your company best practices and policies.
The recommendation you are seeing within your Knowledge Studio workspace are specific to your organization. If you have specific feedback on our recommendations, please send us feedback so we can continue to improve them.
Click on the icon that appears on the bottom right corner of each recommendation and fill out a quick survey to send us feedback.
Updated 17 days ago