How I Use AI Tools in My Developer Workflow Without Losing My Edge
By Harshil Patel · · 4 min read · AI, Productivity, Tools
AI coding assistants are everywhere. Here's how I integrate tools like Claude, Gemini, and Ollama into my workflow while still sharpening my core skills.
The AI Integration Question
Every developer is asking the same question: how much should I rely on AI? After months of experimenting with Claude, Gemini, Whisper, and local Ollama models, I've found a balance that works.
Where AI Shines
Boilerplate generation — Setting up Express routes, database schemas, and utility functions. AI handles the repetitive structure so I can focus on business logic.
Code review — Having an AI scan my code for edge cases and security issues before pushing to production has caught bugs I would have missed.
Documentation — Turning my messy code comments into proper JSDoc or README files in seconds.
Where I Draw the Line
Architecture decisions — AI can suggest patterns, but understanding WHY you choose a monolith vs microservices requires experience and context it doesn't have.
Client communication — No AI can replace the nuance of understanding what a client actually means vs what they say.
Debugging production issues — When something breaks at 2 AM, you need to understand the stack trace yourself.
My Setup
- Claude for complex problem-solving and code architecture
- Gemini for quick lookups and API exploration
- Ollama (local) for sensitive client code I can't send to the cloud
- Whisper for transcribing client call recordings into action items