AI Is Changing My Job (And That's the Point)

Last month I refactored a Laravel codebase - 78 Actions, 3,000+ tests - using an AI agent running in a loop. The kind of systematic migration that would have taken weeks of tedious, error-prone work. It ran overnight.

It also failed in ways I didn’t expect, invented categories I didn’t ask for, and confidently reported completion on tasks it hadn’t touched. I wrote about the gotchas in detail. The point isn’t that AI didn’t work. It’s that making it work required me to be better at directing it than I was at doing the work myself.

That’s the reality nobody in the “AI is taking our jobs” discourse wants to engage with.

The Debate Is Exhausting

You’ve seen the takes. “AI will replace developers by 2025.” “AI is just autocomplete with better marketing.” “Learn AI or become obsolete.” “AI can’t do real work.”

Both camps are wrong, and both are useless.

The doomers assume AI capability means human irrelevance. It doesn’t. The dismissers assume current limitations are permanent. They aren’t. And everyone’s so busy defending their position that they’re missing what’s actually happening.

AI is a tool. It’s changing what work looks like. The people who learn to use it well will outcompete those who don’t. That’s it. That’s the whole thing.

What’s Actually Happening

I use AI every day. Not as a novelty. As infrastructure.

Different models for different tasks. Claude for complex reasoning and code architecture. GPT-4 for certain content tasks. Smaller, faster models for quick lookups and formatting. The skill isn’t “using AI” - it’s knowing which model, which prompt structure, which verification steps, for which problem.

When I’m refactoring code, I’m not typing less. I’m directing more. The work shifted from “write this function” to “specify exactly what this function needs to do, how to verify it works, and what failure looks like.” That’s harder than writing the function. It’s also faster, once you’re good at it.

The prompt engineering fundamentals aren’t optional knowledge anymore. They’re core competency. Not because AI does the thinking - but because the quality of your direction determines the quality of output.

AI amplifies the gap between good and mediocre. A developer who understands architecture, testing, and system design can direct AI to produce exceptional work at speed. A developer who doesn’t will produce the same mediocre work, just faster. Or worse - confidently broken work they don’t know how to verify.

The Real Threat

It’s not AI. It’s the competitor who figured this out six months before you did.

The agency that delivers in two weeks what used to take six. The developer who ships features while you’re still scoping. The business that iterates faster because their technical partner isn’t manually doing what could be automated.

This isn’t hypothetical. It’s happening now. And the gap is widening.

What Adapting Actually Looks Like

It’s not “learn AI” - that’s meaningless. It’s specific shifts in how you work.

Understanding model strengths. Claude reasons well but can be verbose. GPT-4 is good at certain structured outputs. Smaller models are fast and cheap for simple tasks. Knowing which to reach for - and when to switch mid-task - is a skill that compounds.

Building verification into everything. AI is confident. Confidently wrong, sometimes. The Ralph loop taught me that “tests must pass” isn’t a success criterion - it’s an invitation to find the cheapest path to green. You need to specify HOW, not just WHAT. And verify at every step, not just at the end.

Treating prompts as code. Version them. Test them. Review them. A prompt that works today might drift tomorrow. The discipline you apply to production code applies here too.

Integrating AI into workflows, not bolting it on. The wins don’t come from occasionally asking ChatGPT a question. They come from AI being embedded in how you research, draft, refactor, review, and ship. It’s infrastructure, not a party trick.

Staying accountable for the output. AI generated it, but your name’s on it. If you can’t verify the work, you can’t ship the work. That means understanding what the AI produced well enough to catch when it’s wrong.

For Business Owners

You’re trying to figure out whether AI is real value or expensive noise. Fair question.

Here’s one example: presenting ideas used to be expensive. A concept needed mockups, maybe a prototype, possibly a demo environment - all before you knew if the direction was right. Now, ideas can live and breathe early. AI can generate visuals, draft interfaces, build working prototypes, and iterate on feedback in hours instead of weeks. You can explore three directions before committing budget to one.

That’s not about replacing designers or developers. It’s about de-risking decisions. The ideas that survive early exploration get proper investment. The ones that don’t get killed cheaply. That’s better for everyone.

What to look for in partners and hires:

• They can explain how they use AI, specifically. Not buzzwords - actual workflow.
• They’re faster than they were two years ago, and can show why.
• They talk about verification and quality control, not just speed.
• They’re honest about what AI can’t do.

Red flags:

• “AI will handle it” with no explanation of oversight.
• Speed promises with no mention of how quality is maintained.
• Dismissing AI entirely (“we don’t need that”).
• Treating AI as a cost-cutting measure rather than a capability multiplier.

The value isn’t cheaper work. It’s more work at the same quality, or better work at the same speed. If someone’s selling you AI as a way to pay less for the same output, they’ve missed the point.

For Developers

You’re anxious. That’s reasonable. Here’s what actually helps.

Start with your existing workflow. Where do you lose time? Boilerplate? Research? Documentation? Code review? Pick one. Find an AI tool that addresses it. Get good at that before expanding.

Learn prompt engineering properly. Not “tips and tricks” - the actual discipline. Structured prompts, few-shot examples, chain-of-thought reasoning, output formatting. This is the skill that transfers across every model and every use case.

Build verification habits now. The developers who thrive with AI aren’t the ones who trust it most - they’re the ones who verify fastest. Learn to write prompts that produce checkable output. Learn to spot when AI is confidently wrong.

Stay technical. AI doesn’t replace understanding - it amplifies it. The developer who understands database indexing can direct AI to write performant queries. The one who doesn’t will ship slow queries faster. Architecture, testing, system design - these matter more now, not less.

Ship something with AI this week. Not a toy project. Real work. Make a mistake. Learn from it. The gap between “I’ve read about AI” and “I use AI” is enormous, and you only close it by doing.

The Point

Adapt or be replaced sounds dramatic. It’s not meant to be. It’s just true, and it’s always been true.

Version control replaced developers who couldn’t adapt. So did automated testing. So did cloud infrastructure. So did CI/CD. At each shift, people worried about obsolescence. The ones who learned the new tools didn’t become obsolete - they became more valuable.

AI is the same pattern, faster. The window to adapt is shorter. The advantage for early movers is larger. But the principle is identical: learn the tool, direct it well, stay accountable for the output.

I’m not worried about AI taking my job. I’m using it to do my job better. That’s the whole point.

---

We help businesses and development teams integrate AI into their workflows - not as hype, but as practical capability. If you’re figuring out where to start, get in touch.