🎼 The AI Orchestra: Rovo's Dual Role as Conductor and Instrument in the Agentic Future

[#1]

The evolution of AI in the Atlassian ecosystem has been a fascinating journey, from simple chat to agents powered by our internal context. The most recent and exciting shift is Rovo's role in the emerging Agentic AI landscape, where it operates simultaneously as both the conductor of the workflow or a high-fidelity instrument within a larger ensemble.

This dual role is critical, and it’s founded on the strength of the Teamwork Graph.

 

Context is King

Before Rovo can conduct or play, it needs to know the sheet music. We established that the journey moved from Standalone Agents (general LLMs) to Contextual Agents (Rovo connecting to internal knowledge).

The Teamwork Graph provides this context, mapping the relationships between every Jira issue, Confluence page, Bitbucket commit, and user across your organization. This semantic layer is why Rovo's insights are precise, and its actions are targeted. It understands how work is connected, not just what the words mean.

The Agentic Future: Conductor and Instrument

The true power of Rovo lies in its flexibility to act at different levels of the AI workflow, enabled by the Forge platform and the broader ecosystem standards like the Model Context Protocol (MCP).

Role 1: Rovo as the Conductor 🎻

As the orchestrator, Rovo is responsible for managing complex workflows that span multiple domains. This is achieved by leveraging third-party agents via Forge Actions.

In this scenario:

• A user prompts Rovo with a goal (e.g., "Draft a release note for the features completed in Jira project X and get legal approval").

• Rovo acts as the conductor, breaking the goal into steps.

• Step Delegation: Rovo executes a custom Forge Action that calls a third-party AI service (e.g., a specialized content generation agent for drafting, or an external compliance agent for review).

• The final, verified output is then posted back into Confluence or Jira.

Rovo conducts, ensuring that specialized, best-in-class AI services are used for the parts they excel at, with all information flow respecting the Atlassian trust boundary and permissions.

Role 2: Rovo as a High-Fidelity Instrument 🎺

When the primary workflow lives outside the Atlassian ecosystem (e.g., in a terminal, a major coding IDE, or another foundational AI service like Gemini or ChatGPT), Rovo becomes the most valuable instrument: the authoritative source of truth for all work-related context.

Thanks to integrations like the Atlassian Rovo MCP Server, third-party services can programmatically call Rovo to perform work in the Atlassian world.

In this scenario:

• A developer uses the Gemini CLI to ask a question.

• Action Invocation: Gemini recognizes that the question requires internal data and sends a request to Rovo (the instrument).

• Rovo instantly fetches, summarizes, or acts on data from Jira, Confluence, and Bitbucket, and passes the accurate, context-aware result back to the Gemini CLI.

This powerful interoperability ensures that regardless of the entry point, whether you're using a third-party IDE or another AI platform, Rovo is the essential tool for accessing, understanding, and executing tasks against the unique Atlassian context of your team's work.

This is the true spirit of Agentic AI: a world where tools collaborate seamlessly to enhance human productivity.

Our learnings

The "Conductor" approach with the Google AI API was straightforward. Setting up a Forge Rovo Agent (rovo:agent) and connecting Actions was quick and easy.

The challenge emerges when it comes to fine-tuning and coordinating the prompts. Keep in mind that Rovo will process the response from Google a second time. This is actually a benefit when the external response is structured data (for example, JSON) and needs to be rewritten in a human-readable format. The downside of this approach is that you have to define remote URLs, which means the app loses the "Runs on Atlassian" badge since data leaves the native environment.

Using Rovo as an "instrument" means you open up the Atlassian world to your AI tool of choice. I used it while coding with the Gemini CLI to read and update work items in Jira. Furthermore, I used it to change documentation directly in Confluence. The range of actions it can take is really impressive. You can also connect it with ChatGPT (at least outside of the EU), which brings these powerful capabilities to non-technical users like Product Owners or Marketers.

So what's your thoughts on Rovo being the conductor or an instrument?