I’ve been using AI coding assistants (currently Claude Code, but I’ve also used Copilot Chat and Cursor) as part of my daily workflow for the past few months. As someone managing an engineering team, I need tools that actually work — not toys that demo well but fall apart in production.
Here are 10 tips that genuinely save me hours every week. These work with most AI coding assistants — the examples show CLI-style prompts, but adapt the approach to whatever tool you prefer.
1. Use Your AI Assistant for Code Review Pre-screening
Before I review my team’s PRs, I run them through my AI assistant first:
git diff main..feature-branch | claude "Review this diff for bugs, security issues, and style problems"
This catches ~70% of the issues I’d normally flag, so my actual review focuses on architecture and logic — the stuff that matters. Between this and the other review tricks, I save a good chunk of my review time. (I wrote a full breakdown of this workflow in How I Use AI to Review Code Faster.)
2. Generate Test Cases from Requirements
Instead of writing tests from scratch:
AI assistants are surprisingly good at thinking of edge cases you’d miss. I usually keep 80% of what it generates and tweak the rest. This alone probably saves me a couple hours a week.
3. The “Explain Like I’m Onboarding” Trick
When inheriting legacy code:
I used this when taking over a legacy data pipeline. What would’ve taken a week of code reading took an afternoon.
4. Batch Refactoring with Context
Don’t refactor file by file. Give your AI assistant the full context:
It understands cross-file dependencies better than you’d expect.
5. Write Migration Scripts, Not Migrations
Instead of writing database migrations manually:
The downgrade part is key — most AI assistants will include it if you ask explicitly.
6. Debug Production Logs Faster
Paste a stack trace and ask:
The ranking by probability part forces structured thinking instead of a dump of possibilities.
7. Generate Documentation That Developers Actually Read
The curl example is the secret — developers will actually test it.
8. Use It as a Rubber Duck (But Smarter)
When stuck on architecture decisions:
It won’t make the decision for you, but it’ll surface tradeoffs you haven’t considered.
9. Sprint Planning Prep
Before sprint planning:
This saves the “wait, this ticket is way bigger than it looks” surprise mid-sprint.
10. Automate the Boring Commit Messages
git diff --staged | claude "Write a conventional commit message for this change. Be specific about what changed and why."
Small thing, but good commit messages compound. Your future self (and your team) will thank you.
The Meta-Tip
The biggest productivity gain isn’t any single trick — it’s building the habit of asking “can AI do this for me?” before starting any repetitive task. And how you ask matters — good prompt engineering patterns make a huge difference in output quality.
Half the time, it can. The other half, it can’t — but you only lost 30 seconds asking.
What AI coding tricks have you found useful? I’m always looking for new workflows.
You might also like
- Cursor vs Claude Code in 2026: Which AI Coding Tool Should You Use?
- Prompt Engineering for Developers: 7 Patterns That Actually Work
- How I Use AI to Review Code Faster (Without Missing Bugs)
📦 Free: AI Code Review Prompt Pack — 10 prompts I use on 15+ PRs/week.
Newsletter: One practical AI workflow per week, plus templates I don’t publish here. Subscribe →