Just the other day, I passed Atlassian’s Rovo Fundamentals certification (ACH-800, here’s the link to the actual credential if you need proof)—20 out of 20, for whatever that’s worth. That said, the course itself didn’t teach me much.
I’d been using Rovo every day for months (years, actually) before I ever looked at the study material. The centerpiece framework in the course, TCREI (Task, Context, References, Evaluate, Iterate), is really just basic prompt engineering with an Atlassian label. Anyone who has spent real time with ChatGPT, Claude, or Gemini already knows the drill. The certification didn’t teach me the skill; it just validated what I’d already built through trial and error.
So why use Rovo at all when the other tools are honestly better in a lot of ways? Because Rovo already has the context. Ask Gemini or Claude a question about your project out of the box and it can’t answer. It has never seen your Jira tickets or your Confluence pages. Rovo sits right inside both, and pulling across live tickets and docs is what turns a generic answer into something you can actually use.
That clicked for me when I switched teams at work. I’d been at the company almost two years, knew the people, and had worked in an adjacent part of the business the year before, but my old team owned one piece of the system and the new team owned the rest. I wanted the fastest path to getting up to speed, starting with a dense assessment document full of technology I had very little context for.
The knowledge wasn’t missing; it was just scattered across Confluence pages and Jira tickets that nobody had time to stitch together by hand. Instead of booking hours of people’s time to walk me through the basics, I used Rovo to connect the dots: distilling the docs, explaining unfamiliar components simply, and working through follow-up questions on my own.
After sanity-checking my understanding in a few quick messages, it took just one focused meeting with our lead business analyst to validate what I’d learned. My ramp-up went from weeks of shadowing to a few days of targeted reading, and I hit the ground running. It kept the homework on me to read and learn, which is where it belongs.
The bigger test came later with end-to-end test planning. On our program, that work normally sits with business analysts and QAs, and it’s a heavy lift.
I pointed Rovo at every story written for every feature, along with the acceptance criteria inside each one and other documentation, and had it draft an end-to-end plan with full coverage. It mapped out the test data each scenario needed and traced every scenario back to the exact stories it covered so we could prove nothing slipped through the cracks. Then it generated the test cases themselves as Jira tickets: step-by-step grids traced back to each story’s live status, so if a story wasn’t ready yet, the tester knew to skip that step.
The plan lived in Confluence, linked to every case across in progress, complete, and blocked. When new scenarios came up, I didn’t start over. I built a reusable Rovo agent that spins up new test cases the exact same way in minutes.
I did the heavy lifting upfront to get the structure and draft cases built, and then served it up to our BAs. They’re closer to the features than I am, so they reviewed and refined all of it, without having to spend days building the scaffolding from scratch. I now run an end-of-day status read the same way, flagging which functional areas are slipping against our target dates and where we might need to shift resources.
Which brings me to the question I actually want to ask: why is nobody talking about this?
Part of it is the hype cycle. All the oxygen right now goes to ChatGPT and Claude, and to the claim that software development as we know it is ending. For a brand-new company building greenfield, I mostly agree. Why would you start with anything other than agents?
Most of the industry doesn’t work that way, though. Established companies have decades of development and mission-critical systems running on traditional processes, and the Atlassian suite is where a huge share of that work lives. Those companies don’t change their practices on a dime, especially when software supports the business rather than being the product. The problems are still big and genuinely specific to them, and the practices are solid; they just don’t need to be bleeding edge. Tools that live in that world rarely get the spotlight of the frontier labs.
I’ll be honest about the limits, too. The harness around Rovo isn’t as sharp as what I’m used to from Claude. It doesn’t always answer consistently, and Atlassian’s implementation can be frustrating. If I could run Claude fully integrated into Jira and Confluence, I might not use Rovo at all.
Right now, I can’t. Rovo is the one that’s actually inside the content, respecting the permissions and access rights that come with a company’s proprietary information—and in day-to-day delivery, that turns out to matter more than raw model quality.
So I’m genuinely curious: is anyone else out there using Rovo this way? Not just the chat window for a quick summary, but agents, plans, and real delivery work. I almost never hear about it, and I’d love to compare notes.
AI assisted with the writing of this post.