The honest version: I didn’t. I’ve been running one since 2015. I just never put a name on the door until this year.
That name is ByteShip. It’s managed WordPress hosting and maintenance, which in 2026 sounds a little like announcing you’ve opened a fax repair shop. AI can build a website in about ninety seconds. Why would anyone pay a human to look after one?
Fair question. Here’s my answer.
Launching was never the hard part
Every website I’ve ever built launched fine. Launch day is the easy day. Everything is fresh, nothing has drifted, the client is happy, the plugins are all on their current versions and getting along.
Year two is the hard part. Year two is when the payment gateway quietly stops firing receipts. When the contact form starts landing in spam because nobody ever set up authenticated sending. When a plugin update breaks a layout on one template that nobody looks at except the three people who were about to sign a contract. When PHP hits end of life and the host sends an email to an address that belonged to an employee who left in March.
None of that is a building problem. It’s a watching problem. And AI did not solve the watching problem. It made it bigger, because now there are more sites to watch.
AI didn’t reduce the number of websites. It multiplied them.
The assumption baked into “why hosting, in this economy” is that AI replaces websites. It doesn’t. It replaces the barrier to making one.
So the number of sites goes up. And the average site is now built by someone who has never heard of a database, doesn’t know what a nightly backup is, and has no particular reason to. That’s not a criticism. That’s the whole promise of the tools, and it’s a good promise. But it produces a very specific gap: a lot of infrastructure owned by people with no infrastructure habits.
I like the AI site builders. I’m not competing with them. The AI that built it doesn’t do watch shifts.
The job of a website changed, and it got more important
Here’s the part I actually find interesting, and the reason I stopped treating ByteShip as a back-office function and started treating it as a business.
For twenty years a website was a destination. You did SEO so people would arrive, and once they arrived you tried to convert them. Traffic in, action out.
That model is eroding. People increasingly ask an assistant instead of clicking a blue link. Which sounds like bad news for websites, and for a certain kind of website it is.
But watch what the assistant is actually doing. It’s reading. It’s pulling facts about your hours, your pricing, your services, your team, your policies, and it’s answering on your behalf. Your site stopped being the place people go and became the place the answer comes from.
That’s a promotion, not a demotion. A destination can be mediocre and still work, because a human will squint and scroll and figure it out. A source of truth cannot. If your structured data is wrong, the wrong answer gets repeated. If your site is slow or blocked or malformed, you don’t get cited at all. Somebody else does.
Nobody is going to notice that happening. There’s no alert for “an AI decided you weren’t worth quoting.” That’s exactly the kind of slow failure that maintenance exists to catch.
The objection I take seriously
I should say the uncomfortable thing out loud, because I think about it a lot.
If assistants answer more questions directly, some sites will get less traffic. Real traffic, real leads, gone. Publishers are already living this. Anyone whose entire model was ranking for a question and monetizing the click is in genuine trouble, and no amount of clean metadata fixes that.
I don’t think that’s most of my clients. A consulting firm, a municipality, a nonprofit, a growth-stage founder: for them the site was never a traffic machine. It was proof they exist and evidence they’re credible. That job survives. It might even get more valuable, since the assistant needs something authoritative to point at.
But I’d be selling you something if I said this shift is upside for everyone. It isn’t. It’s upside for people whose site is a source of truth and downside for people whose site is a toll booth.
What ByteShip actually does
The pitch is short. Automation handles the routine work. A human verifies the results.
Updates run on a schedule. Backups run nightly. Uptime is watched. Security is patched. That part isn’t special, and I’d be lying if I told you it was. Software does most of it, and software is good at it.
What software is not good at is knowing that the update technically succeeded but the homepage now renders wrong on mobile. Or that the backup completed but hasn’t been test-restored in eight months. Or that the form submits but the email never arrives. Green checkmarks are easy to produce. Somebody has to actually look.
That’s the whole company, honestly. The human side of the automation. Every site also gets future-prepped for AI search, which means structured data, clean metadata, an llms.txt file, and a site that assistants can actually read and cite.
Why now, after eleven years
Because for eleven years this was the invisible thing I did for design clients under the Today Creative umbrella, and I finally noticed I’d been treating the most durable part of the business as an afterthought.
Design work is project work. It ends. Maintenance doesn’t end, which is the part everyone finds boring and the part I’ve come to think is the actual asset. It compounds. It builds trust slowly. It’s unglamorous in a way I’ve made peace with.
I didn’t build a WordPress hosting company because of AI. I built one, kept it quiet, and then AI made the argument for me.
ByteShip is at byteship.com. If you’ve got a site you’re not entirely sure anyone is watching, that’s usually the sign.