Rovo agents are AI “teammates” that live inside Atlassian Cloud. They help teams find information, learn from existing content and perform tasks. They follow the same permissions as your users, so answers and actions stay within what each person is allowed to see and change. For an organisation, this means fewer hand‑offs, less context switching, and faster outcomes across IT, HR, finance, product, and beyond.
1. Start with one small, high‑value job
Pick a weekly, low‑ to medium‑risk task. Examples: scope‑creep checks on Jira epics, docs‑to‑update suggestions, or drafting first replies to common support questions.
2. Understand the difference between Rovo-Agent Identity and Agent Scenarios
Think of Agent Identity as the agent’s job description and personality (“who it is and how it behaves”) and Agent Scenarios as the specific tasks it can perform (“in this situation, do X using Y information”). A good pattern is:
- Identity: stable, high‑level and rarely changed.
- Scenarios: concrete, focused, and safe to iterate on as you learn.
3. Onboard your Rovo-agent like a teammate
Write the Rovo-agent Identity as if onboarding a colleague – role, tone, and guardrails. Avoid vague “helpful assistant” descriptions.
4. Narrow the knowledge per scenario
Scope to the minimum a well‑trained human would use (e.g., one space like IT‑Support‑KB; only how‑to and troubleshooting). Less noise, fewer hallucinations.
5. Stay assistive before you automate
Start with drafts and suggestions. Only enable stronger actions (updates/creation) once results are consistently reliable.
6. How to build trust in your Rovo-agents
Validate with Real-World Scenarioser
Test using actual work, like real Jira work items for example, not hypothetical examples. Involve subject-matter experts and Jira administrators early to ensure the solution works in real operational conditions, not just on paper.
Maintain a consistent set of 10–20 test cases and use them as a baseline when iterating. Improvements in one test case can unintentionally reduce quality in another test case, so track results consistently to avoid regressions
Let the agent improve its own instructions
Ask it to explain decisions, correct the target behavior, then have it rewrite the scenario instructions. Iterate and paste improvements back into Rovo Studio.
How to know if you’ve built a good Rovo agent?
- Each agent performs one job reliably from start to finish.
- You have provided clear instructions on exactly what the agent should do in various situations (scenarios).
- You have scoped the agent’s knowledge to only what is relevant for solving the specific task at hand.
- You have established regular quality tests to ensure the agent’s performance remains high over time.
This guide has provided you with the initial building blocks, but the potential of Atlassian Rovo goes far deeper. Are you ready to explore how Rovo Agents can truly streamline your specific workflows?
Download the complete guide to building Rovo agents.
We have gathered our expert insights into a comprehensive whitepaper: “A complete guide to unleash your potential with Rovo-agents” This is the roadmap for those looking to transition from simple tests to advanced automation that delivers tangible business value.
Click here if the form does not appear on your device
You must be logged in to post a comment.