My Rovo Use Cases in Jira Service Management

[Julian Beyer]

Hi everyone,

I've been exploring using Rovo for automating repetitive tasks and wanted to share some practical examples. I’m very interested to hear what’s working well for you!

Here are my automations that currently use Rovo in some capacity. My personal impression is that they have made our support workflows faster and more consistent:

• Ticket Criticality Assessment:
  When a new ticket is created, Rovo analyzes its content and suggests a priority (with an explanation) in an internal comment.
• Re-Assessment After Customer Comments:
  If a customer comments on a ticket, Rovo re-evaluates the priority and updates the priority suggestion if needed, with reasoning provided.
• Resolution Summaries:
  On issue closure, Rovo generates a structured summary (category, summary, rationale, resolver, outcome) as an internal comment.
• Status Review and Synchronization:
  Rovo regularly reviews bugs and feature request tickets we receive in Support. It regularly analyzes related development items, and suggests statuses and comms as needed to keep everything in sync.
• Customer Inactivity Handling:
  After a period of customer inactivity, Rovo helps double-check if the ticket was indeed resolved and logs the result, helping us fine-tune automated closure and reporting.
• Handling of Comments on Closed Tickets:
  When a closed ticket receives a new comment, Rovo is involved in deciding whether to revive the ticket or ignore the comment, escalating to humans for edge cases.
• CSAT Analysis:
  When a CSAT rating is entered, Rovo analyzes the outcome, identifies the root cause, and sends a summary to key staff.

• Stale Knowledge Base Page Reminders:
  In addition to the regular Knowledge Base team doing their reviews, Rovo checks our knowledge base pages for staleness, suggests specific changes and initates the update process.

I’d love to hear from others:

• What kinds of AI-powered automations have you built in Jira Service Management? Are there any that stick out as especially valuable or challenging)?
• I make sure to keep humans in the loop as much as possible and hand off the final decision to them, especially for critical or uncertain cases. Do you have cases where you trust Rovo enough to let it decide on an action that could have direct customer impact?
• Do you use many custom Rovo agents in your rules or are the default Rovo agents enough for you?

If anyone is interested in the specifics of how these automations are built, I’m happy to explain how the rules are built and walk through the setup process.

[Nicolas Bourdages (Ubisoft)]

We're experimenting with a custom rovo agent (not a virtual support agent, that's too rigid for our needs) to chat with customers to define their needs and point them towards the appropriate documentation or create a JSM request on their behalf with the information gathered during the conversation. The agent guides the customers in clarifying their needs and providing the necessary parameters. It's extremely promising!

[Alex Stillings]

Jullian,

Great list of really great use cases for any user support queue. I'd love to learn more on some of these and see how you are leveraging them. Would you be will to post one of the instructions and setup? Or we can connect directly.

[Julian Beyer]

Sure, here's a simple example. This one shoots a quick email to key staff about new CSAT ratings (this is not for deep dives, just to keep your finger on the pulse):

Trigger: Field value changed (any changes to the field value Satisfaction)

Action: Use Rovo Agent (you can use the default "Customer Insights" agent or duplicate it and shorten its instructions a bit)

Analyze this ticket’s CSAT outcome: {{issue.url}}

Pinpoint likely root cause.
State who fixed it and how.
Be ultra-brief—think telegram, pay-per-letter.
Use bullet points.
No need to repeat the satisfaction rating.
If the fix was unsatisfactory, briefly note possible reasons.

Action: Send customized email

🎯 <b>New CSAT Rating Received!</b><br>
📝 <b>Ticket:</b> <a href="{{issue.url}}">{{issue.key}}</a><br>
🗂️ <b>Summary/Ticket Name:</b> {{issue.summary}}<br>
🌟 <b>Rating:</b> {{issue.satisfaction.rating}}/5 ⭐<br>
🙋 <b>Reporter:</b> {{issue.reporter.displayName}}<br>
🧾 <b>Organizations on this ticket:</b>

{{#issue.organizations}}

{{name}}

{{/issue.organizations}}
<br>
💬 Customer comment: <a href="$linktosatisfactionreport">Click here to view it in the satisfaction report</a><br>
🤖 <b>Rovo summary:</b><br>
<blockquote style="border-left: 4px solid #ccc; margin: 8px 0; padding-left: 12px;">
{{agentResponse}}
</blockquote>

[Yanqing Lin]

Julian, 

    You have built a great Rovo agent for JSM. I am also quite interested in your solution to automate the workflow. 
    Do you mind sharing your agent ( removing all your PII data) please? Or direct message.

cheers

Yan

[Julian Beyer]

I prefer to use multiple specialized agents. So I have not one just agent, but different agents for different rules.

You can use one of the default agents as a baseline, then cut down their instructions to the parts that are actually relevant for the automation. Since it doesn't cost anything to duplicate agents and edit their scenarios, you can organize this any way you think makes sense. To get started I can recommend [Workshop] Let's build a Rovo Agent: JSM Triage and (Handout) Let's Build a Rovo Agent: Jira Triage Automation - Confluence

For example here is the agent for Status Review and Synchronization. That one helps us observe and sync the statuses of support requests with the development work going on in the background.

It is invoked by a rule on a scheduled trigger. The rule then runs on all open Bugs and Feature requests in our JSM project.

Behavior:

You are the Bugs & Features Status Arbiter, an agent invoked via automation. You are invoked to review and update current Bugs and Feature requests in $JSMproject. Your tone is methodical, formal, and analytical.

Scenario instructions: 

You are the Bug & Features Status Arbiter, an agent invoked via automation. You are regularly invoked by automations to review current Bugs and Feature requests in $JSMproject. For each item:

1. Examine the related $R&Dproject items of the $JSMproject item you are given.
2. Analyze the statuses and comments of those $R&Dproject items.
3. Form an objective judgment on whether a status change for the $JSMproject item is warranted, based on the $R&Dproject items' progress and discussion.

You can find the status definitions for Bug and Feature work types and how they map to $R&Dproject statuses in $page
The process that R&D uses is explained in $page

- Your tone is methodical, formal, and analytical. Keep all responses short.
- If no status change is needed, do nothing and do not respond.

You can then give it more specific instructions whenever you call on it in an automation, like asking it for reasoning or a draft for comms.

Since the agent can't take any actions by itself when invoked in automation you are still in control of what happens in the end. I use its {{agentResponse}} to inform these actions:

1. Changing the status between different variants of Open/work in progress/pending (I do not let it close/resolve something by itself)

2. Leaving an internal comment explaining the reasoning for the status change.

3. Leaving an internal comment with a suggestion for comms:

Hey [~accountid:{{issue.assignee.accountId}}], there have been new developments in this case. I think you should send an update to the customer. Here's a suggestion:{{agentResponse}}