Use the virtual agent in your help center

Allow customers visiting your help center to quickly get the information they need without having to manually search for articles or request forms. The virtual agent uses generative artificial intelligence to search across your linked knowledge base spaces, summarizes that information to answer customer questions, and remembers context for follow-up questions.

How it works

1. Customers enter their question or search term into the search bar on your help center. Read more about the help center.

2. Using Atlassian Intelligence, the virtual agent summarizes information from your connected knowledge base to respond. Read more about Atlassian Intelligence answers.

   • If the customer’s search term is too vague or doesn’t match any information in your knowledge base, the virtual agent asks them to rephrase their question.

3. After an answer is provided, customers can ask follow up questions. The virtual agent remembers previous questions from that conversation, and uses customer context (for example, the location set on their user profile) to provide more accurate responses.

4. With each answer, customers are offered a thumbs up or thumbs down as a way to gather feedback about the quality of each answer. This feeds into the Helpfulness score provided on your Performance page.

Customer conversations are stored for 30 days for analytics purposes, but the details of these conversations won’t be visible to anyone. Keywords from customer conversations may be anonymized and presented on the performance page to help you improve the performance of the virtual agent.

View and improve the performance of the virtual agent in your help center

You need to be a site admin or org admin to view the performance of your virtual agent in the help center.

To see the performance of the virtual agent in your help center:

1. From your service project, select Channels, and then Portal.

2. Select your avatar in the top right hand corner.

3. Select Performance.

4. To get a sense of overall traffic and how the virtual agent is performing, look at:

   • Customers: The total number of customers who asked at least one question.

   • Questions: The total number of questions asked by customers.

   • Answered: The number of times an Atlassian Intelligence answer was provided in response to a customer’s question.

   • Helpfulness: The percentage of answers rated by customers that were rated as helpful. Keep in mind that you might have a helpfulness score of zero if your customers aren’t rating any answers.

5. To improve the performance of your virtual agent in the help center, look at Create helpful articles.

   • You’ll see groupings of keywords that are being entered by customers but aren’t returning responses from the virtual agent. Use these keywords as a guide to create (or edit) knowledge base articles to make sure that customers searching with these keywords get the answers they need.

   • Create helpful articles may not appear on your Performance page until you’ve had enough unanswered questions to group keywords.