Why I Went All-In on AI-First Engineering

About six months ago I stopped treating AI coding tools as a novelty and started building my entire workflow around them.

I’d been using Claude and various AI tools casually. Code completion here, a chat query there. Useful, but incremental. Then I started building actual workflows around AI: Claude Code for daily development, MCP servers connecting AI to our internal tools, custom agents handling code review and test generation.

The difference wasn’t incremental. It was transformational.

What changed

Three things shifted when I moved from “using AI tools” to “building AI-first”:

1. Velocity compounded

It wasn’t just that individual tasks got faster. The compound effect of AI assistance across an entire sprint was staggering. Code generation, test writing, PR descriptions, documentation, debugging. Each one 2-3x faster adds up to a fundamentally different pace of delivery.

2. Quality went up, not down

This surprised me. I expected AI-assisted code to need more review, not less. But Claude Code catches patterns I miss, suggests edge cases I hadn’t considered, and writes tests I wouldn’t have bothered with. The code isn’t just faster, it’s more thorough.

3. The boring work disappeared

Every engineer has tasks they dread. Boilerplate, migration scripts, config files, documentation updates. AI handles these with zero reluctance and consistent quality. That means engineers spend more time on the interesting problems — architecture, design, performance, user experience.

What I’m using daily

My current stack for AI-powered engineering:

• Claude Code — My primary development companion. I use it for everything from greenfield features to debugging production issues.
• MCP Servers — Connecting Claude to our internal APIs, databases, and deployment tools. This is where it gets powerful — AI that can read your actual system state.
• Custom AI Agents — Automated code review, test generation, and deployment verification. These run in CI/CD and catch issues before humans even look at the PR. (I wrote about building these agents and my MCP setup in separate posts.)

The Staff Engineer angle

As a Staff Engineer, my job isn’t just to write code. It’s to multiply the effectiveness of the entire engineering organisation. AI-first engineering is the biggest force multiplier I’ve seen in my career.

I’m not talking about replacing engineers. I’m talking about giving every engineer on the team a tireless, knowledgeable pair programmer that never has an off day.

The teams I work with ship faster, with fewer bugs, and with better documentation than they did a year ago. That’s not a marginal improvement. It’s a step change.

Getting started

If you’re an engineering leader thinking about AI adoption, here’s my advice: don’t start with a tool evaluation. Start with a workflow audit.

Look at where your team spends time on low-creativity, high-repetition work. That’s where AI will have the biggest impact. Then pick one workflow, instrument it with AI, and measure the difference.

That’s been my experience, anyway.