Rovo Automations?

[Colton_Kosicek]

Hi

I'm noticing  that ROVO is not great at handling "complex" tasks that have multiple ACTION outputs.  In general the ideas can very easily be handled by ROVO but the ACTIONs functionality seems to be lacking if I have a single agent to help.

Do i need to create multiple agents to take care of each individual step of my actions so it fully operates?

Example: Review this feature - rewrite it's description and acceptance criteria (different fields) the system will duplicate the same information in both fields or miss AC field. 

If i ask it to create user stories after this it will create just description without Acceptance criteria and forget it's parent is the same features it's using to create content from.  ANY insights or timelines for when this will become more general available/capable would be appreciated.

[Bernandina Ortiz · Accepted solution]

Hi @Colton_Kosicek 

I have been experimenting with creating Rovo agents, and I completely agree with you. From what I have learned, they perform best when focused on a single, specific task. By leveraging automation rules, you can then orchestrate multiple agents to work together and accomplish more complex tasks.

[Colton_Kosicek]

Have you been able to understand how to get the formatting to be consistent and clean?

[Bernandina Ortiz]

The formatting of the instructions to the Rovo agent?

[Colton_Kosicek]

1. this is how it decides to repeat ac

   a.. Acceptance Criteria

         1. details of acceptance criteria

               A.  Given........

               B.  When............

               C.  Then..........

[Bernandina Ortiz]

You can actually tell it how to structure the output. Try giving it an example. The reason that it default to that style is because it is what it has been programmed to do by default. 

[Dr Valeri Colon (Connect Centric)]

@Colton_Kosicek Rovo Actions can fire multiple steps, but they’re still limited in how reliably they chain updates. In complex flows, agents often overwrite fields or lose context. The most dependable approach today is splitting the workflow across smaller agents or using one agent with very strict, step-by-step instructions. Atlassian is improving multi-action reliability, but no public timeline yet.