Best practices for using AI in Atlassian products
15 min
Beginner
By the end of this lesson, you'll be able to:
- Considerations while searching in Jira using Atlassian Intelligence
- Optimize your Jira search prompts for better results
- Describe the best practices to get accurate results
Atlassian Intelligence results are based on existing data
Importance of access and up-to-date data
- Atlassian Intelligence will only find what exists If you don't have permission to access it or something doesn't exist, then Atlassian Intelligence won't display any output relating to your prompt. It can't process something that doesn’t exist.
- Keep Atlassian objects such as Jira work items or Confluence pages up-to-date Maintaining current and accurate information in your space and objects is important. Using AI to search for answers in Confluence yields the best results when your Confluence site is populated with detailed, complete, and up-to-date content. Regularly adhere to best practices across all products. Ensure that you archive items when possible and keep statuses current. In Jira Service Management, the quality of AI-generated responses will be affected if your knowledge base articles are outdated or not regularly updated.
Be cautious about sharing confidential company information
Even though Atlassian and its LLM partners guarantee not to share your data for further AI training purposes, we still recommend avoiding sharing personal, confidential, or sensitive data.
AI technology is undeniably powerful and excels when paired with the appropriate mindset. Both Atlassian and its LLM partners prioritize data security by ensuring that all transmitted information is encrypted. However, it is essential to recognize that certain types of data should not be shared, even with these protective measures in place.
Searching work items in Jira using Atlassian Intelligence
Atlassian Intelligence works best in the following scenarios:
✔ Search for fields and values that exist in your Jira project.
✔ Use specific fields and values in your query that can help narrow down your work item search.
✔ Searches in projects you have access to.
✔ Query is in English.
✔ Query is translatable to JQL. Since Atlassian Intelligence converts prompts to JQL code, inputs containing keywords that can be translated to JQL can provide better results.
Optimize AI prompts for better results
Persistence is key here. Try different things and optimize along the way. Observe how your AI responds and tweak the prompts based on its performance. Think of this as an ongoing dialogue where you're continually refining the conversation.
👉 Initial query: Imagine a situation where you, as Alana's manager, need to review the work items assigned to her. You might start by asking, "Find all work items assigned to Alana". You get over 1,000 results distributed across more than 20 pages! Going through so many pages would make it quite challenging to locate the specific information you need.
👇 Here's an example and results of “find all work items assigned to Alana” in Jira.
👉 Refined query: To view Alana's work items in the specific project titled "mobile dev," we can refine the query for clarity. By adding the project name, the updated query becomes: "Find all work items assigned to Alana in project mobile dev." With this refined search, you now have only 9 work items to review.
👇 Here's an example and results of “find all work items assigned to Alana in project mobile dev” in Jira.
Using JQL, you can filter your search using the fields available in your project, such as priority, assignee, and work item type, combine multiple criteria, and look for important work item, such as the ones that you’re watching or the ones that are blocked.
Tip: Sort on the column you are searching on. If the column is not already visible, make it appear using columns in Jira. In the example above, the column Project was made to appear so that we know that project mobile dev was matched as mentioned in the criteria.
Using Atlassian Intelligence in Jira Service Management
When creating knowledge base articles for virtual service agents, keep the following in mind:
- AI answers don't extract information from images.
- AI answers perform best with text, not in tables in Confluence.
- AI answers quality relies on the knowledge base's accuracy and clarity.
- A well-structured, up-to-date knowledge base enables better question deflection by the virtual service agent.
👇 Here's an example of where you can give feedback after using AI.
Give feedback
After using any of the Atlassian Intelligence features, you can rate how useful it was to you by selecting the thumbs-up or thumbs-down icon. This helps improve the features for your organization. Please share which parts of the output were helpful and which could be improved. If you notice any inaccuracies in Atlassian Intelligence, let us know in your feedback so we can make corrections.
👇 Here's an example of where you can give feedback after using AI.
How was this lesson?
next lesson
Best practices for writing AI prompts
- The basics of AI prompting
- Write a great AI prompt
- Continue the conversation with follow-up prompts