Core Concepts
AI-Assisted Development
Using AI effectively when building with Webiny.
Overview
AI can accelerate Webiny development when used as a practical assistant for exploring code, scaffolding features, and generating extensions. The most effective approach treats AI like a capable teammate: provide context, define outcomes clearly, review output, and refine.
AI in Development Workflow
A typical AI-assisted workflow:
- Understand the feature or extension point to modify
- Gather context from codebase, docs, and existing patterns
- Ask AI for focused implementation
- Review for correctness and maintainability
- Test locally and iterate
Where AI Helps Most
AI excels at accelerating implementation tasks:
- Creating extension drafts
- Scaffolding Admin UI changes
- Generating GraphQL queries, mutations, resolvers
- Wiring lifecycle-driven integrations
- Creating utility functions
- Producing TypeScript boilerplate
- Explaining unfamiliar code
- Debugging errors and suggesting causes
- Transforming ideas into structured code
Use AI to accelerate implementation, not replace technical decision making.
Where Developers Lead
Developers must control critical areas:
- Architecture decisions
- Where custom logic should live
- Business rule validation
- Security and access control
- Multi-tenant behavior
- Maintainability and upgrade safety
- Edge case testing
AI generates solutions, but developers own the consequences.
Effective Prompting
Good prompts include:
- What you’re building
- Which Webiny part you’re customizing
- Where implementation should live
- Constraints to respect
- Expected output format
Poor prompt:
Better prompt:
Provide Context First
Before requesting implementation, provide:
- Relevant file/folder structure
- Existing code to follow
- Extension point to use
- Documentation excerpts
- Constraints (coding style, upgrade safety, tenant awareness)
Break Into Focused Tasks
AI performs better with smaller, defined tasks.
Instead of:
Break into steps:
- Define data model
- Identify extension point
- Generate UI component
- Add backend logic
- Connect lifecycle events
- Test and refine
Request Platform-Native Solutions
Guide AI toward solutions that:
- Use extension points vs. modifying core
- Follow existing project structure
- Keep logic modular
- Use TypeScript properly
- Remain maintainable
- Avoid brittle shortcuts
Useful instruction:
Review Generated Code
Don’t just check if code compiles. Ask:
- Does this belong in the right place?
- Does it follow project structure?
- Does it introduce duplication?
- Does it respect tenant boundaries and permissions?
- Does it feel like proper Webiny extension?
- Will another developer understand this later?
- Will this survive future upgrades?
Common Mistakes
Avoid these pitfalls:
- Requesting large implementations in one step
- Insufficient project context
- Accepting code without review
- Letting AI invent patterns vs. following existing ones
- Forcing custom logic into wrong system parts
- Treating prototypes as production-ready
Recommended Workflow
Practical AI-assisted workflow:
- Use docs and context to understand platform area
- Inspect existing code before generating new code
- Ask AI for small, focused implementation
- Review against platform patterns
- Test locally
- Refine through short iterations vs. giant rewrites
This makes AI a force multiplier instead of a source of code requiring rewrites.
Best Practices
- Context before code - Gather relevant information first
- Small iterations - Break work into reviewable chunks
- Pattern alignment - Match existing Webiny patterns
- Security review - Check permissions and tenant isolation
- Test thoroughly - Validate edge cases and production behavior
- Maintain standards - Keep code readable and upgradeable
AI significantly improves productivity when used to explore, scaffold, and accelerate. Best results require intentional architecture, structure, and review decisions from developers.
The goal isn’t just generating code—it’s generating code that belongs in your Webiny project and moves it forward cleanly and maintainably.