I've Spent 30+ Years Fixing Computers.
This Year I Let AI Help Me Build a Website.
I've been a working technician for over 30 years — long enough that I remember when "fixing a computer" meant reseating a card or reflashing a BIOS, not untangling somebody's malware infection or explaining why their laptop won't hold a charge anymore. I know hardware. I know troubleshooting. Coding, though, is a different story — and it's a bit more complicated than just "I don't know code."
Truth is, I've built a handful of simple websites over the years. The original Big Mike Computer Services site was one I put together myself — a single page, same colors as my business card. That's actually how I connected with a local web developer: he saw that page, gave me a call, and we struck up an arrangement — he'd build me a proper one-pager, and I'd send him people who needed a decent website done. Over the years I've probably sent him three or four good referrals that way.
So I'm not a total non-coder. I just never kept up as websites got more complex — and at some point they went from "something I could tinker with" to "far more involved than I felt comfortable tackling solo."
That's the gap I've been closing this year. When I decided to develop the idea for this site, FMC247, I'd heard about using AI to help build websites. After a bit of research, I settled on Claude — it seemed to offer the best value and came with the strongest recommendations from what I could find. I figured I'd write down what that's actually been like, because I suspect a lot of small business owners are wondering the same thing I was: can someone who isn't a full-time coder actually pull this off with AI doing the heavy lifting?
Short answer: yes, but not the way I expected.
The Wins
It catches things I'd never catch myself. A while back we ran a full audit of every article and support page on this site — 26 articles, 6 support pages at the time — and found broken internal links, broken footer links, and a CSS bug where certain buttons and callouts were missing their accent color site-wide. I'd looked at those pages a hundred times and never noticed. A script caught it in minutes.
Fixes get done properly, not just patched. When we found broken link problems, we didn't fix them page by page — we used various scripts to build out an entire internal link network connecting related articles, then built a proper pillar page to tie it all together. That's the kind of structural work I would never have known to prioritize on my own; I would've just fixed the broken links and moved on, and it would have taken me forever.
It respects what I actually know. I was firm from the start: I do not claim to be a "certified" anything, and I won't be pretending like I've personally used every product I write about if I haven't. The AI doesn't fight me on that — it just writes it straight. If I haven't used something, we say so and link an equivalent instead of faking it.
The Frustrations
Website files drift. More than once we started editing a version of a page that wasn't actually what was live on the server — usually because I'd made a small change myself, or because the copy we were working from was a session or two old. Lesson learned: always start from a fresh download of what's actually live before touching anything, when working with someone (or some thing) else — like AI.
Some bugs are sneaky. There's a mobile display bug I'm still chasing — a whole "Most Popular" section of article cards just disappears on narrow screens while the section around it keeps its full height, like the cards fell through a trap door. We know roughly what's causing it. Hunting it down has taken patience.
AdSense doesn't care how good your content is. I assumed if I were to build solid, from-the-heart content, the ads would follow. Instead, one of my three sites got flagged for "low value content" and I hit a review-attempt limit that locked me out of resubmitting for weeks. Meanwhile another of my sites sailed through. Same effort, same care, very different outcomes — which tells me this is as much a numbers/volume game as a quality one.
I also expected AI to write things I could paste in without reading, but that's not how I use it. Every article gets read by me line by line before it goes live — I catch phrases that don't sound like me, numbers that need verifying, claims that need softening to "unverified" until I've confirmed a price or a spec myself. The AI drafts fast, but I still own what actually gets published.
The other surprise: how much of this work isn't writing at all. It's fixing broken canonical tags across dozens of pages at once. It's untangling a malformed affiliate link. It's figuring out why AdSense's consent-messaging settings work the way they do. None of that is glamorous, and none of it is stuff I'd have learned or even wanted to learn fixing hardware for three decades — but it's exactly the kind of systematic, "why is this actually broken" thinking that troubleshooting computers trained into me in the first place. That basic premise transfers into webbuilding as well.
Where This Is Going
I'm not retiring on this yet. But piece by piece — fixing what's broken, building what's missing, being honest about what I do and don't know — this site is slowly becoming something "real" (as in personal and vulnerable) instead of just something that exists. If you're a small business owner wondering whether AI tools are worth the learning curve: for me, the answer has been yes, definitely — just as long as I stay in the driver's seat.
📧 BigMikeNanaimo@gmail.com