How close is Rovo Dev to fully autonomous ticket resolution in real-world projects?

Mahanth Prudhvi P

Hi everyone,

I recently watched a demo of Rovo Dev where it appeared to handle an entire Jira issue autonomously. Once a ticket was created or assigned to the agent, it seemed to understand the requirements, prepare the necessary code, and close the ticket with very little to no human involvement.

The demo was impressive, but I'm curious about how this translates to real-world software development.

For those who have used Rovo Dev in production or active projects, I'd love to hear your experiences:

- How much of the end-to-end development workflow can Rovo Dev realistically handle today?
- Can it genuinely take a Jira issue from assignment to a code-ready state with minimal human involvement, or is this mostly achievable in controlled demo environments?
- What types of work is it best suited for (bug fixes, small enhancements, refactoring, documentation, test generation, etc.)?
- Where does it still require significant developer intervention or review?
- How reliable has the generated code been for production use?

I'm also interested in understanding the ecosystem and prerequisites required to achieve a workflow similar to the demo:

- Which Atlassian products are typically involved? Is Jira, Bitbucket, and Confluence additional products required
- Can it work effectively with repositories hosted on GitHub, GitLab, or Azure DevOps, SVN or is the experience primarily optimized for Bitbucket?
- How does it fit into existing CI/CD pipelines? Can it prepare code and seamlessly participate in the pull request, build, test, and deployment workflow?
- Are there any specific integrations, automation rules, or Rovo capabilities that need to be configured to enable this level of automation?

Finally, based on your experience, how close is the current reality to the product demonstrations? Are teams successfully using Rovo Dev to handle a significant portion of development tasks autonomously, or is it currently better suited as an AI development assistant that still requires regular developer oversight?

I'd really appreciate hearing about your real-world experiences, lessons learned, and any limitations you've encountered. Thanks in advance!

Jason Krewson

I'm actually looking into this myself right now with Cursor and below is a workflow I've been slowly moving toward to determine whether it's possible. At the moment, I'm waiting on approval to connect Cursor, so I'm a bit blocked until that comes through. Once that does I will work on Claude as well.

Possible Workflow

• My application throws an error.
• Rovo creates a Bug work item.
• The Bug is assigned to Cursor.
• Cursor reviews the Bug details.
• Cursor analyzes the relevant application code.
• Cursor proposes a solution and I review/approve it.
• Cursor implements the fix.
• Cursor commits and pushes the updated code to the repository.
• Cursor notifies me when it's complete.
• I review, test, and approve the changes.
• Cursor closes the work item.

Not sure how much of this is possible yet, but I feel this is what the future can look like.

As for the topic, Rovo, when you say "Once a ticket was created or assigned to the agent, it seemed to understand the requirements, prepare the necessary code, and close the ticket with very little to no human involvement." it sounds like the agent read the requirements and then updated the ticket with the recommended code in the given example requiring manual effort to update that code in the app and then let Rovo agent know it can close the work item, is that correct?

If so I think that is a good use of these agents right now, I mention this to people here all the time (and in my monthly newsletter), you can assign an agent to your work items and they can even help you code, it's a great start. 

To get something like this fully automated, it depends on several integrations and connections. I think a lot of it comes down to the company you work for and how comfortable they are with that level of AI access and automation.

My company is a little strange when it comes to AI. They're actively pushing AI adoption, but at the same time I've had AI related initiatives shut down without a clear explanation. About a year ago I was trying to integrate Copilot with Jira and that effort got stopped.

That said, Cursor falls under a different team so maybe I'll have better luck this time. 😂

Anyway, great post!