When Atlassian introduced Rovo, all the talk was about how good the model is, how well it is integrated, and of course, the vast array of tasks you can delegate to it. But it is not only about Rovo’s capabilities, it is about how to make the most out of it.
In this blog post, we will walk through the common mistakes teams make when using Rovo and how to get more consistent and reliable results.
Your information inside Confluence & Jira isn’t well structured or up to date
When asking Rovo to create drafts and pull data from the sources you manage inside or outside your Atlassian suite, it is important you give it up to date and well structured information. And the reason for this is quite straightforward: Rovo can only work with what it has. If the source is outdated or inconsistent, it might reproduce those issues in your draft.
Always make sure to update your sources and archive any outdated pages or projects as Rovo doesn’t know what’s obsolete unless you remove or flag it. Additionally to further help Rovo, you can leverage labels to provide extra context and facilitate the pulling of information.
Your prompts are vague and generic
The more specific you are, the better the results. If your prompt is a bit vague with little to no details, you end up with generic results and a long, unnecessary chat. Rovo fills in the gaps for you but not always in the way you expect. Instead of a tailored output, you often get a generic structure that still requires heavy editing or full rewriting.
Your prompts, especially the first one, should provide Rovo with the bigger picture. In it you will answer some key questions. What is the goal? Who is my audience? How do I want them to experience my page? Etc. Once you’re clear on these, include clear instructions for the output and formatting.
| Bad Prompt | Good prompt |
| “Reformat this text into a documentation” Too vague. AI doesn’t know the audience, goal, or how to structure content, so the output is generic and requires heavy editing afterwards. | “Structure this documentation page for end users. The goal is to make it skimmable and actionable. Include an AI-generated summary at the top, organize the key sections with headings and include a table of contents” Specific about audience, purpose, structure, and desired experience, guiding AI to produce an initial draft. |
Your prompts are not actionable
When writing prompts, the verbs you use guide the AI on exactly what you want it to do. When prompts are passive or unclear, Rovo tends to return descriptive or loosely structured text instead of something you can directly use.
Prompts that use action words such as “Create”, “build”, “draft”, and “search” produce clearer, structured, and more actionable output.
| Bad Prompt | Good prompt |
| “Make an onboarding plan for new hires for 90 days” The prompt is not action oriented and lacks key details on what the plan should include. | “Draft a 90 day onboarding plan. Using information from the following Confluence pages [Confluence page link]. Highlight the intro and key takeaways in an info panel. Create three headings for the first week, month and 90 days” Actionable words are used with extra details provided for a clear structure. |
Your prompts are too complex
AI struggles when asked to do everything at once from one single prompt. When creating a draft from scratch or editing your page’s existing text/formatting, make sure to break down your requests so it focuses on one element at a time rather than the entire page. Without this, you risk getting a single draft where AI loops on similar outputs, leaving little variation regardless of follow-up prompts.
Additionally, if you want some specific macros to be included within your page, you should mention them within your prompts. Rovo can’t know whether your bullet points are supplementary information that should be included within expands: You should tell it.
| Bad Prompt | Good prompt |
| “Structure the entire page with headings, panels, expands, and actions” This prompt is broad and doesn’t specify which content belongs in which section, forcing AI to guess the structure and often resulting in a disorganized or inconsistent page layout. | “Add a panel at the top summarizing the page’s purpose and key takeaways. Make it clear and skimmable.Turn the three bullet points in the second section into expands” Each prompt is specific which reduces AI guessing and ensures the output aligns with the intended page layout and user experience. |
You don’t experiment enough with your prompts
When it comes to prompts, there is no one magical formula. If you rely on one prompt style, you end up locked into one type of output structure. This limits creativity and often leads to repetitive content patterns.
Even the slightest of changes can provide greatly different results. This is why it is important to experiment and iterate. Testing different approaches lets Rovo explore multiple formats, giving you more flexibility and creativity in how your outpout is structured across your Atlassian suite.
You don’t review generated content
Everything starts with us humans and finishes with well… us (for the time being at least).
Even after a detailed and productive discussion with Rovo, the final human review is crucial to make sure everything is on point before sharing your page. AI can give you structure, and the initial draft with proper formatting, but it can’t replace your tone, understanding of the topic, and of course, judgement.
Here is what to look for when reviewing your co-created page:
- Overall page structure:
This is the initial scan. Check whether the right headings are used, white space, dividers, and column placement. AI may rearrange content in a visually dense way.
- Text accuracy in formatted sections:
Make sure all macros still contain the correct content and are placed where intended. Check that the specific sections you asked Rovo to insert within macros such as expands, panels, and others haven’t been altered. AI may change text along the way (not that often though).
- Key details from other sources:
When pulling information from multiple sources like a Jira project for example, make sure to check that you have entered the right project name, timeframe, etc. Also, always double-check that what’s in your page matches the source.
Getting value from Atlassian Rovo isn’t just about using the tool, it’s about how you work with it.
Across all the issues we’ve covered, a clear pattern emerges: Rovo performs best when it operates in a clean, structured, and intentional environment. That includes well-maintained Jira and Confluence data, clear and actionable prompts, and a thoughtful approach to iteration and review. For more practical tips, examples, and best practices around Confluence, Jira, and Atlassian Rovo, make sure to take the blog tour.
