ROVO provides inconsistent results

[Sowmya]

Hi,

I’ve been using ROVO to generate sprint summaries, including capacity (as defined in the sprint goal) versus committed story points. I’m also asking it to provide my commitment based on roles, details on completed stories, RAG status based on defined criteria, and accomplishments derived from completed work, and then update all of this on a Confluence page.

While the descriptive summary generated by ROVO aligns well with the sprint information, I’m noticing that the quantitative data being fetched is not always consistent. 

Could you please suggest how to address or troubleshoot this?

Thanks,

[Arkadiusz Wroblewski]

Hello @Sowmya 

https://support.atlassian.com/jira-software-cloud/docs/use-atlassian-intelligence-to-search-for-work-items/

What you are asking Rovo to do here is really two different things at once: write a sprint summary and also calculate or interpret sprint metrics. That second part is where you need to be more careful.

Atlassian’s own docs say Rovo output can vary, and their Jira documentation also says Rovo is less useful when you want it to analyze work items into summaries or other data representations. So for values like commitment, completed story points, capacity, or a RAG based on fixed rules, don’t treat the generated output as the source of truth.

For that part, use Jira’s native reports or a clear JQL-based data set first, then let Rovo turn that verified data into the written Confluence update. The Sprint Report and Velocity Report are the better reference points for those numbers.

What usually helps is tightening the scope a lot. Use the exact sprint name, board or filter, the estimate field, the definition of commitment, what counts as done, and the exact RAG logic. I’d also split this into two steps instead of one large prompt: first get the exact numbers, then ask Rovo to write the summary from those results.