Every developer I know uses AI now. The ones who say they don't are either lying or writing code the slow way out of spite.
The real conversation isn't whether to use it. It's the massive gap between what AI-powered web development actually looks like versus what LinkedIn influencers want you to believe. One camp says you can vibe-code an entire production app in an afternoon. The other camp says AI writes garbage and real developers don't need it. Both are wrong, and the truth is way more useful than either hot take.
I've been building web applications for over a decade. Enterprise platforms at Bell Canada. Financial systems at Taulia before SAP scooped them up. Ecommerce and live events at Sofar Sounds. Frontend teams at Wells Fargo. I've seen enough production code to know what "good" looks like and what "this will break at 2am on a Saturday" looks like. AI is part of every project now. It makes the work faster. It does not make the work easier. Those are very different things.
Design Phase: AI Generates, I Curate
AI can spit out layout concepts and design directions in minutes. Feed it brand guidelines, competitor sites, and project goals. You get a stack of starting points before your coffee gets cold.
Starting points. Not finished designs.
AI design output gravitates toward safe and familiar, because that's literally what it trained on. The average of everything. Fine for brainstorming. Not enough for a site that needs to convert visitors or stand out from the twelve competitors who also used AI to generate their layouts.
The real work is knowing which concepts have legs and which are polished dead ends. That instinct comes from years of building interfaces that real people actually use. A SaaS dashboard and a local bakery landing page need fundamentally different things. AI doesn't make that call well on its own.
Keep what works. Kill what doesn't. Push the survivors further than AI would go alone. The design phase takes less time, but the output is still original and strategic. Not a template with a fresh coat of paint.
Code Generation: AI Writes the Bricks, I Design the Building
This is where AI saves the most time. Also where the experience gap matters most.
Boilerplate code is AI's sweet spot. Component scaffolding, data fetching patterns, form validation, responsive grid layouts. The structural plumbing every project needs but nobody's writing a conference talk about. AI handles it in seconds.
Architecture is a completely different conversation. AI doesn't know your ecommerce site needs to handle 10,000 people during a flash sale. It doesn't think about service layer patterns so you can swap CMS providers later without rewriting your entire frontend. Code splitting, caching strategies, how your component library holds up when someone adds 30 pages next quarter. None of that lives in a prompt.
At Bell Canada, the work involved bilingual Next.js applications with headless CMS integrations deployed on AWS. That same architectural thinking applies to a five-page small business site. The scale changes. The standards don't.
Content: AI Drafts, I Make It Sound Like a Human Wrote It
Most clients need help with website copy. This is where AI has genuinely changed what smaller budgets can afford.
AI drafts initial copy based on the client's business, audience, and goals. Headlines, product descriptions, about page content, blog outlines. A reasonable first draft of all of it, fast.
Reasonable and good are not the same word. AI copy is bland, repetitive, and deeply in love with words like "leverage" and "innovative" and "solutions." It writes in a voice that belongs to nobody. Every brand ends up sounding like the same corporate ghost wrote their website.
So every piece gets edited for voice, accuracy, and specificity. The fluff gets cut. The copy ends up sounding like the actual business, not like a LinkedIn post. And everything gets fact-checked, because AI will confidently state something completely wrong with the energy of a guy who is absolutely sure he's been to that restaurant before.
AI gets the draft to 60%. Editing gets it to 100%.
Testing: AI Does the Boring Part Well
Writing tests is one of those things developers know they should do and almost always skip when deadlines show up. Every. Single. Time.
AI shifts that equation. Unit tests, integration tests, edge cases. It's particularly good at thinking through the weird scenarios a human would skip. What happens when the input is empty? A string with 10,000 characters? An API returning a 500 error on a Tuesday for no discernible reason?
The test scaffolding and a lot of the cases come from AI. The review is mine. There's a real difference between tests that pad a coverage number and tests that actually catch things before they break in production. Knowing that difference is kind of the whole job.
SEO: AI Crunches, I Decide
Search engine optimization involves a lot of pattern analysis and data processing. AI was basically built for this.
Keyword research, search intent analysis, meta descriptions, structured data markup. AI processes competitor analysis faster than any human could, spotting gaps and opportunities in minutes. The analytical work is genuinely impressive.
SEO strategy is a different muscle. Which keywords are worth targeting given the client's domain authority? Can we realistically rank for this term? Local search or broader? These calls require understanding the client's business, their competition, and how search actually works right now. Data alone doesn't get you there.
AI gives me better data to work with. The strategic calls are still mine.
"If AI Is Doing the Work, What Am I Paying You For?"
Fair question. Short answer: me.
You're paying for the person who knows what to ask for, what to keep, and what to throw away.
Anyone can prompt AI to generate a website. Tutorials are everywhere. But the output without experienced oversight is the development equivalent of fast food. It looks like a website. It technically functions. It's not built to last, not optimized, not accessible, and it will create problems that cost more to fix than doing it right would have.
The same standards I held at Wells Fargo and Bell Canada apply to every project I take on. Code gets reviewed. Architecture gets planned. Performance gets tested. AI just means that level of quality ships faster, at a price point that works for businesses that aren't Fortune 500.
You're paying for the developer. AI is one of her tools.
The Bottom Line
AI doesn't replace developers. It makes experienced developers faster. The gap between an AI-assisted junior developer and an AI-assisted senior developer is just as wide as it ever was. Probably wider, honestly. Knowing what to do with AI's output matters a lot more than knowing how to prompt it.
If you want modern tools and someone who's been around long enough to use them well, let's talk.
