Rovo Agent trining

[David Alonso]

Hi people!

A question regarding the long-term behavior and improvement of AI agents. Is there any formal training or learning process for AI agents over time? In other words, will an AI agent behave the same on day 1 as on day 1000, or is there a way for it to improve based on feedback, usage, or guidance?

If such a process exists, I’d appreciate any recommendations on how I can support the ongoing training of my AI agent and what I should communicate to my users so they can contribute effectively. My goal is to make sure the agent keeps delivering the highest possible value as it evolves.

Thanks in advance for any insights or best practices!

[Rebekka Heilmann _viadee_]

Hi @David Alonso 

like @Melbrin NGOUILOU said, Atlassian has announced that Rovo Chat now has memory. All changes regarding testing & improving Agents are still in developement. See What's New in Rovo Agents: Your Bi-Monthly Update - Atlassian Community for the latest news and updates.

So: it's in the works but don't think anything gets delivered this year.

[Melbrin NGOUILOU]

Hello,

Great news! The answer to your question is yes! Atlassian announced at Team Europe 2025 that AI agents can now be trained and improved over time. In Atlassian Studio, you’ll find a dedicated section in the agent configuration where you can review past conversations and apply corrections or adjustments.

These corrections are then used to generate a new version of your agent, which you can safely test in a private environment before redeploying it to your users.

Hope this helps!