What inspired you?

We are living through a revolution where AI writes code faster than humanly possible, but this speed creates a dangerous bottleneck. If AI drives development at 100mph, manual testing becomes the traffic jam waiting at the end of the line. We realized that for the "System of Work" to flow, Quality Assurance wasn’t just about observing the AI revolution, but active participation as well. We built the Test Case Architect to ensure that the speed of AI-generated code is matched by the rigor of AI-generated testing, establishing a continuous "Chain of Trust" from requirement to release.

What does it do?

The Test Case Architect operates as a specialized Rovo Agent integrated directly into the Appsvio Test Management app. It automates test design through a four-step technical workflow:

  • Requirement Interpretation: The Agent analyzes the full context of a Jira Issue - interpreting summaries, descriptions, and comments - to extract explicit and implicit acceptance criteria.
  • Scenario Logic: It applies testing methodologies to generate a matrix of scenarios, ensuring coverage across Happy Paths, Negative flows, and Boundary conditions.
  • Schema Enforcement: The Agent outputs strictly formatted steps that map to the ATM data structure: Action, Data, and Expected Result.
  • Forge-Native Provisioning: Upon confirmation, the Agent triggers Forge invocations to instantiate native Test Case objects and automatically establishes the trace link to the source requirement.

What have you learned?

One of the biggest lessons was that building a Rovo Agent is less about writing code and more about "AI Psychology.". In the world of Rovo, context is currency. Our responsibility wasn't just connecting pipes to Jira, but acting as architects of user intent - anticipating every persona, from the exhausted QA Lead to the detail-oriented Tester. Furthermore, we discovered that prompt engineering isn't a checkbox task, it resembles more of a lifecycle. So, our prompt is like a living codebase that requires constant refactoring to ensure Rovo remains a reliable, evolving team player.

How did you build your project?

Since Appsvio Test Management (ATM) is the first fully Forge-native test management app, we had a massive home-court advantage - we were already speaking Rovo's language. The building process was a focused mix of structure and stress-testing. At first, we tailored the "Base Prompt" to define the Agent's personality: authoritative on quality, strictly structured on format, but flexible on scenario generation. Another critical phase was the "Human vs. AI" stage. Our QA specialists tried to break the Agent with vague requirements and edge cases. This rigorous feedback loop allowed us to refine the logic before merging the Agent directly into ATM and deploying to the Marketplace.

What challenges have you faced?

We’ve certainly fought a battle for determinism. While Generative AI thrives on creativity, software integration requires absolute structural rigidity. Our biggest challenge was that the Agent naturally wanted to "freestyle" the formatting, whereas our system needed a strict schema to turn text into native Jira issues. We had to fight hard to "tame the beast" - relentlessly refining the prompt to force a reliable, unshakeable output structure (separating Action, Data, and Expected Result) without restricting the AI's ability to think critically about the testing logic.

What are the accomplishments you are proud of?

Without a doubt, we are incredibly proud to have empowered QA teams to match the velocity of development without sacrificing quality. By integrating Rovo directly into the native Jira workflow, we’ve proven that AI isn't here to replace testers, but to elevate their work. We’ve successfully shifted the workload: the Agent handles the repetitive documentation and maintenance, allowing the human experts to focus on strategy and complex problem-solving. We are happy to showcase the teamplayer for modern QA teams based in Jira.

What’s next for the Test Case Architect?

We view this project not as a finished task, but as a continuous evolution alongside the Forge platform. We are already iterating now: in the upcoming version of ATM, we’ve embedded a "Create Test Case with Rovo" button directly on the Requirement panel. By putting the Agent right at the tester's fingertips, we will be able to eliminate the need to search for it. Looking forward, we are setting our sights on the Forge LLM component (currently in Early Access Program) to further deepen the Agent's capabilities. We also plan to expand Rovo's scope from just creation to optimization - allowing it to refine and polish steps in existing test cases. Within all of these changes, our goal remains simple: to keep building a "helping hand" that handles the routine load, while giving testers the freedom to apply their expertise where it shines.

Built With

  • forge
  • forge-sql
  • jira
  • jira-cloud-rest-api
  • kysely
  • react.js
  • rovo-agent
  • tanstack-query
  • tanstack-virtual
  • typescript
  • vitest
Share this project:

Updates