luisgomez
← Back to Thoughts

AI in My Dev Workflow: From Skeptic to Bullish

· 3 min read ·
AI Workflow Opinion

I was the guy rolling his eyes at every “AI-powered” tool announcement.

“Great, another buzzword,” I’d think, while manually writing the same boilerplate for the hundredth time.

Then something changed.

The Moment It Clicked

I was staring at a critical feature. Deadline: yesterday. The kind of pressure that makes you question your career choices.

I fed the requirements to an AI assistant. Not expecting miracles. Just… help.

What came back wasn’t perfect. But it was 70% of the way there. And that 70%? It was the boring 70%. The stuff I had done a thousand times. The CRUD scaffolding. The type definitions. The repetitive validation logic.

I spent my energy on the 30% that actually mattered—the business logic, the edge cases, the things that required understanding why we were building this, not just what.

The Problem Was Me, Not the AI

Here’s what I learned: AI code generation isn’t magic. It’s a force multiplier.

What it does well:

  • Generate boilerplate, fast
  • Handle tedious, repetitive patterns
  • First drafts of documentation
  • Exploring API alternatives quickly
  • Finding syntax I forgot

What it doesn’t do well:

  • Understand your business context
  • Make architectural decisions
  • Know your team’s conventions
  • Replace actual thinking

The developers struggling with AI? They’re treating it like an oracle. Asking it to make decisions it can’t make.

The developers winning with AI? They know exactly what to ask for. They review everything. They supervise like a senior dev watching an intern.

My Current Setup

I’m not using AI for everything. But for specific things? It’s a game changer:

  • Documentation: First drafts of README files, API docs, inline comments explaining complex logic
  • Prototyping: Fast MVPs to validate ideas before committing
  • Research: Understanding new libraries, patterns, or approaches
  • Refactoring: AI suggests improvements, I decide if they’re good

The Catch

I always verify. Every single line.

AI hallucinates. AI misses context. AI generates code that “works” but does the wrong thing.

The skill isn’t using AI. The skill is knowing when to trust it and when to override it.

The Real Change

I shipped a critical feature in record time. Three weeks compressed to five days.

Was the AI code perfect? No.

But we closed the deal. We got the feature out. We refined it after.

That’s the point, right? Build, ship, iterate. Not build, perfect, ship never.

AI helped me do that. Not by replacing me, but by handling the stuff that doesn’t need me.

What’s your take? Drop me a message if you’ve had similar experiences—or if you’re still skeptical. I get it. But maybe give it one more shot with a specific, high-value task.