Building Big Apps with AI Coding Tools — Without Making a Mess

TL;DR I use AI coding tools for scaffolding, tests, docs, and boring boilerplate. Architecture and module boundaries stay human‑led. We broke the app into small vertical slices, added guardrails, and shipped faster.

The Real Problem

1. Large applications slow down as complexity grows.
2. Code reviews take longer as dependencies multiply.
3. Once the codebase exceeds the model’s context window, AI becomes unreliable.

I worked on a similar application and learned how to keep a mid‑to‑large system clean as it scales.

The Core Approach

Break the application into small, focused modules (Auth, Payments, Notifications, Search, etc.).

Give each module a clear contract (OpenAPI, AsyncAPI, typed interfaces).

• Boundaries — What belongs inside each module
• Communication patterns — REST, events, or shared types
• Data ownership & consistency — including CDC when needed
• Cross‑cutting guardrails — auth, security, rate limits, uptime
• Testing strategy
• Security testing
• Repo + delivery strategy
• Where to avoid using AI

Where AI Helps (and Where It Doesn’t)

Where AI Helps

• Scaffolding: controllers, services, repositories, DTOs, validation pipelines
• Boilerplate: logging, error handling, config, DI, authorization setup
• Tests: unit tests, contract tests, test data, basic security tests
• Documentation: READMEs, API examples, module guides
• Refactoring: renaming, restructuring, cleanup

Where AI Hurts

• Architecture decisions
• Complex business logic (edge cases, conditions, failure modes)
• Security‑critical components
• Anything involving sensitive code

AI is an assistant — not the architect.

Pros

• Speed: huge boost for repetitive work
• Consistency: cleaner naming and patterns
• Source control: easier merges and commit hygiene with structured prompts
• Better documentation: faster to create and update
• Less time wasted on syntax issues

Cons

• Quality drift: code looks right but isn’t safe
• Contract drift: modules misalign if not enforced
• Shallow tests: AI defaults to happy paths
• Privacy & licensing concerns
• Architecture erosion: quick fixes break long‑term integrity

Without reviews, expertise, contract tests, and security checks, you’ll hit a wall.

My Workflow

1. Strong context + knowledge files (clean system prompt, templates, cleanup routines)
2. One‑page module ADRs for Users, Auth, Payments, Notifications, Search
3. Contracts‑first approach using OpenAPI + AsyncAPI
4. Scaffold with AI — one module at a time
5. Add guardrails (coding standards, formatting, security scans, contract tests)
6. Ship vertical slices — API + UI + tests per module
7. Document continuously — READMEs, API examples, ADRs

Final Thought

AI accelerates the productivity. Architecture protects

Use both intentionally to build fast and build clean.

— Preetam