Summary
Challenge: Convex, an international speciality insurance firm, handles a high volume of internal FinOps queries through Slack and Jira Service Management (JSM). JSM’s Virtual Agent triages most requests automatically, but in around 12% of cases it cannot classify the request and leaves key fields blank – forcing manual intervention and slowing resolution.
Solution: Automation Consultants built a custom exception-handling Rovo Agent that steps in whenever the Virtual Agent cannot classify a request. It analyses the issue description, determines the correct Query Type and Routing, updates the Jira fields automatically and adds a clarifying comment.
Outcome: More accurate routing and faster resolution across the queries that previously needed manual triage. Human agents now spend less time on classification and more time resolving the queries themselves.
Background
Convex Insurance (Convex), an international speciality insurance firm and longstanding Automation Consultants client, has embraced Rovo across its Atlassian toolset – particularly in Jira Service Management. As part of our ongoing work together, we look for opportunities to automate internal processes wherever it adds value, and request triaging proved an ideal fit for Rovo.
Challenge
Convex’s finance operations team was handling a significant volume of FinOps queries, including:
- Where to reallocate excess cash in a policy
- Requests for more information on a payment
- Missing information on a policy
- Incorrect policy status in the finance system
- Discrepancies in the tax or commission due on a policy
Users raised these requests in a dedicated Slack channel, where the JSM Virtual Agent picked them up, triaged them and routed them to the right team. This worked well for most requests – but for around 12%, the Virtual Agent could not identify the correct triage or routing options. Those tickets landed with blank fields, requiring someone to classify them manually before work could begin, which slowed resolution and pulled the team away from the queries themselves.
Solution
We built a custom Rovo Agent to take over whenever the Virtual Agent cannot classify a request.
There are many possible Query Type and Routing combinations. When a user raises a request in Slack, the Virtual Agent interacts with them to gather additional information – but if their responses don’t match a recognised combination, it creates the ticket with the Query Type and Routing fields left blank.
How the request triage and routing Rovo Agent works
When these fields are left blank, an automation rule triggers the Rovo Agent to process the issue description. The Agent analyses the request, determines the correct Query Type and Routing, and the automation rule applies its response to the ticket — updating the fields and adding a clarifying comment so the receiving team has context.
Rovo is well suited to this work. The Virtual Agent is optimised for the initial customer conversation, before a ticket exists; Rovo excels at interpreting the ticket content itself. Once a request is routed, the Rovo Agent also uses the information in the request to suggest a starting point for potential fixes – although it does not implement any changes itself, keeping humans in control of resolution.
Benefits
The custom Rovo Agent now handles all failed routings – around 12% of total FinOps requests. As a result, Convex is seeing a significant reduction in manual intervention and more accurate triage and routing, freeing the team to focus on resolving queries rather than classifying them.
