A discussion started when @John Funk shared a frustrating experience building a Forge app with AI assistance. What should have been a simple "no-code" experience turned into hours of debugging, deployment issues, deprecated APIs, and conflicting guidance.
The underlying question quickly became:
How close are we to non-developers successfully building production-ready apps with AI?
The Promise
AI-powered app builders promise that anyone can:
- Describe an idea
- Generate an app
- Deploy it
- Start creating value
And to be fair, several Champions have successfully built working apps this way.
The Reality
The challenge is reliability. Several Champions reported issues such as:
- Deprecated APIs being suggested
- Incorrect deployment guidance
- Missing permissions
- Cross-product installation requirements
- Repeated debugging cycles
- AI confidently recommending fixes that didn't solve the problem
As one Champion summarized:
The iteration count is the problem.
The app often gets close to working, but reaching a production-ready result can require significant troubleshooting.
What Experienced Builders Are Doing
A recurring theme was that experienced builders often pair AI with traditional development tools. Common approaches included:
- Claude + Forge MCP
- GitHub Copilot
- VS Code
- Version control systems
- Dedicated testing environments
AI helps accelerate development, but developers still rely on tooling that supports:
- debugging
- rollback
- testing
- code review
- version management
Where Rovo Studio Fits Today
The discussion revealed an important distinction.
- For simple apps, Studio can dramatically reduce complexity.
- For more advanced scenarios, many users still find themselves needing development knowledge and external tooling to create with Forge.
The Product Team Response
One positive outcome from the discussion was direct engagement from Atlassian's Rovo Studio team. They confirmed they are actively exploring:
- App checkpoints and rollback
- Version history
- Repository synchronization
- Better app quality validation
- Improved handling
These were some of the most frequently requested improvements from Champions.
Champion Takeaway
The conversation ultimately landed somewhere between excitement and realism. AI can absolutely help non-developers build useful apps today. But successful app building still benefits from understanding:
- permissions
- deployment
- testing
- governance
- debugging
The bigger question may not be whether AI can build apps. It's whether AI can eventually handle enough of the development lifecycle that users spend more time solving business problems and less time troubleshooting generated code.
Based on this discussion, we're getting closer—but we're not there yet.
