Different outputs using the exact same prompt (same user)

[Georgia Maniati]

I'm receiving different outputs using the exact same prompt in completely fresh chats within minutes of each other. 

Given the level of inconsistency, I cannot refine my prompts to get a desired result. 

Is there a roadmap item addressing Rovo's non-deterministic behavior?

[Shawn Doyle - ReleaseTEAM]

Hi @Georgia Maniati 

That's the nature of AI.  While I suspect it will get better over time, I don't know of a detailed roadmap that describes or defines when.

[Georgia Maniati]

Thanks for the reply.

I understand variability in phrasing or tone is expected with generative AI, but this goes deeper. I'm seeing different content and different sources retrieved from Jira and Confluence, using the same prompt in a clean environment.

When the same input yields different outputs at the retrieval level, prompt engineering becomes ineffective, and reliable usage is impossible. For Rovo to support enterprise-grade functions, there must be a baseline of repeatability. AI can be made deterministic for enterprise contexts; models like OpenAI's support seeding or other techniques to reduce randomness.