How we ran a client's whole website and CMS build with Claude

We built a client's whole website and content CMS by directing Claude, and the substance is in the scope, not the novelty. Structured content collections, brand-voice copy across multiple vendor and article pages, an editable image system that sizes images for the web automatically, schema so search and AI systems can read the pages, and cross-linking between them. For a business commissioning web work, the point is that an AI-directed build can deliver the whole stack, not just a faster first draft of a template.

So this is a case study about what an AI-directed website build actually delivers, and where the limits are. It is one of the builds we did by directing Claude, and it sits with the rest in the Business Case Studies collection, alongside how we rebuilt a Shopify store into a custom theme with Claude Code. We describe the client only as a Perth business in the energy and resources technology sector, because client work stays anonymous until the client says otherwise.

The problem with how web work usually goes

Most owners who have commissioned a website before know the pattern, and it is not a happy one. The build is slow and expensive. The result looks right at handover. Then the content starts to decay, because changing anything means going back to whoever built it, and the copy was baked into the pages rather than held somewhere editable. A year later the site is out of date and nobody on the inside can fix it without another invoice.

The root of that decay is usually structural. When content is hard-coded into pages, the people who own the business cannot maintain it, so it freezes at the state it was handed over in. The site ages not because it was built badly but because it was built in a way that only its builder can change. That is the real cost of a normal handover, and it is the part that an owner feels long after the initial bill is paid.

So when we approached this build, the question was not just how to make a good-looking site quickly. It was how to deliver a site the client could actually keep current after we stepped away. That shaped every decision, and it is the part of the scope that matters most for a business deciding what to commission.

The scope an AI-directed build delivered

The build covered the whole stack, and the scope is the substance here. Structured content collections, so the content lives in editable records rather than baked into pages. Brand-voice copy across multiple vendor and article pages, written to a consistent standard rather than filled in as placeholder text. An editable image system that sizes images for the web automatically, so the client is not manually resizing files to keep the site fast. Schema so search engines and AI systems can read the pages properly. And cross-linking between pages, so the site works as a connected structure rather than a set of islands.

The point for a business commissioning web work is that this is the full build, not a template fill with the gaps filled faster. Structure, copy, images, search readiness, and internal linking are all things a normal web project charges for separately and often skimps on. An AI-directed build delivered them as one coherent piece, directed by one operator. That is a different proposition from "AI wrote the first draft," which is how most people imagine AI involvement in web work. The AI did not just speed up a step. It made it feasible for one person to direct the entire stack to a real standard.

The editable collections are the part that outlasts the handover. Because the content is held in collections rather than welded into pages, the client can edit it without coming back to us. That is the structural fix for the decay problem: the site does not freeze at handover, because the people who own it can change it. It is the single most valuable thing in the scope, and it is invisible at launch and decisive a year later.

What was AI and what was judgement

The honest framing matters, because the value was not "AI made a website." Plenty of tools make a website. The value was that one operator could direct the whole build to a real standard, and that still took direction, judgement, and brand understanding. The copy had to be in the client's voice, which means understanding the client's voice well enough to direct it and to know when the output was off. The structure had to make sense for how the client's content actually worked. The schema and the cross-linking had to be right, not just present.

None of that is automatic. Directing Claude through the build removed the need to hand the whole thing to a development team and wait, but it did not remove the need to know what good looked like at every step and to correct the build toward it. A vague brief would have produced a vague site, the same way a vague brief produces a vague anything. The build was good because the direction was specific, and the direction being specific was the human contribution that no amount of tooling replaces.

So the realistic account is that the tooling collapsed the cost and the timeline of delivering the full stack, and the judgement decided whether the full stack was any good. Those are two different things, and conflating them is how people end up disappointed. The AI made the scope reachable for one operator. The operator's judgement made the scope worth having.

What a business owner can take from it

If you have commissioned web work and found it slow, costly, and brittle after handover, the gap worth knowing about is that an AI-directed build can deliver the whole stack, structure, copy, images, search readiness, and internal linking, with the content held in editable collections you can maintain yourself. The brittleness after handover is a structural choice, and it can be built out. The site does not have to freeze the day you take it over.

Directing that build to a real standard is the work, and it is where the result either holds up or does not, because the tooling makes the scope reachable but the judgement makes it good. We built this client's whole site and CMS this way, anonymised here to sector and city, by directing Claude through the structure, the copy, the images, the schema, and the linking, as one coherent build.

What the client actually owns afterward

It is worth being concrete about what the client owns after a build like this, because ownership is where most web projects quietly fail the buyer. In a normal handover, the client owns a finished site they cannot really change, which means they own the appearance of control without the substance of it. The moment they need a new vendor page, a corrected detail, or an updated article, they are back to commissioning work, because the content was built into the pages and only the builder can safely touch it. They paid for an asset that depreciates the day they take delivery.

Because we held the content in editable collections, the client owns something different: a site whose content they can actually maintain. A new vendor page is a new record in a collection, not a development job. A change to the copy is an edit to a field, not a ticket to a builder. The structure, the schema, and the image system keep working underneath as the content changes, because they were built to hold content rather than to be the content. That is the difference between owning a photograph of a business and owning the business, and it is invisible until the first time the client needs to change something.

This matters commercially beyond the one build, because it changes the relationship after handover. A client who can maintain their own site does not resent coming back for the things that genuinely need expertise, a structural change, a new capability, a redesign, because they are not coming back for trivia they should be able to do themselves. The editable foundation makes the ongoing relationship honest: we are paid for real work, not for being the only people with the keys to a site the client supposedly owns. Building it that way is more work up front and a better deal for everyone after, which is the trade worth making.

The lesson generalises past websites. Any deliverable a vendor hands over can be built either as a frozen artifact only they can change, or as something the owner can maintain, and the second is almost always worth the extra care. Ask of any commissioned build whether you will own the thing or just a picture of it, because that question, asked before the work starts, decides whether the asset keeps its value or quietly decays the day it lands.

If you want this kind of automation built into your business, that is what Ortopylot does. Start a conversation at ortopylot.com.