Turn support tickets into self-service answers.

How to Use the Knowledge Base Agent in HubSpot to Turn Support Tickets into Help Center Articles. The knowledge base agent reads closed support tickets to find what a knowledge base is missing, then drafts edits to existing articles and new articles for a team to review and approve. The demo follows BikeTools as it keeps its help center up to date. All suggestions from the knowledge base agent appear in one place, the Suggestions tab. Each suggestion shows its estimated impact and how often the topic comes up in conversations; for example, one article came up 108 times. The agent suggests two kinds of changes: Update article, which improves an existing article, and Create article, which fills a gap in the knowledge base. Every edit includes a reason. The agent checks whether each article is still accurate. When new tickets contradict what an article says, it flags the outdated information and suggests a correction. It also reviews how each article reads, looking for improvements in clarity, AEO, branding, and voice and formatting. A tooltip explains why the agent suggested a change. Users decide which changes to accept or dismiss. Edits can be accepted and published all at once with Accept all and publish. Accepting edits does not use HubSpot Credits. The agent also finds questions the knowledge base does not answer yet. In the BikeTools example, customers repeatedly ask which tools to buy for basic repairs, and no article covers it, so the agent suggests a new article through a Create article suggestion. Source conversations links to the tickets that led the agent to recommend the article. Generating a new article with Generate article costs 200 HubSpot Credits. The agent writes a full draft built around the questions customers actually ask. The draft can be edited in the existing editor, then published with Accept and publish or saved for later. Each published article gives customers and Customer Agent more accurate answers to draw from, which deflects more tickets and reduces time spent answering repeated questions.