One working day is enough to take a business brief to a published website with a working contact form. I do not mean a demo on somebody’s laptop or a template with the logo swapped, I mean a real site on a real address that the business keeps. I can write that with a straight face because the method is the one that built the site you are reading, and over the past few weeks I have timed it, unit by unit, into a running order that fits between nine in the morning and six in the evening.
This piece walks through that running order hour by hour. It opens a series about a one-day build programme I am putting together, but you do not need any interest in the programme for the schedule to be useful. If you have ever owned a website project and wondered where the months actually went, the answer is in here.
The building moved. The deciding did not.
One day is possible now for a reason that has nothing to do with typing speed. Code was never the real bottleneck on a small business site. Waiting was. A developer slot next month, and then a two-minute change sitting in a queue for three days. What has changed is that an AI coding agent, in my case Claude Code working on WordPress through a live connection rather than through admin screens, does the building the moment a decision is made. The gap between deciding something and seeing it on a page has collapsed from days to minutes.
That collapse changes the arithmetic of a whole project. When every decision costs a week of turnaround, you batch your decisions, the batches queue, and the project stretches across a quarter. When a decision costs minutes, the entire chain of decisions fits in one sitting. Nobody in this format rushes. Nothing waits either, and the waiting is where the months used to live.
The running order
Here is the day as I have designed it, with what exists at the end of each block. Breaks sit between the blocks and the room is capped at twelve people, each building their own company’s site rather than a practice one.
| Time | Block | What exists afterwards |
|---|---|---|
| 9:00 | Brief to sitemap | A page list and one agreed conversion path |
| 10:15 | Wireframes | Three layout directions, one chosen and developed properly |
| 11:45 | Homepage build | The design system, live on a real page |
| 1:45 | Content | A full article, imagery, metadata and internal links |
| 3:30 | Full-site assembly | Every remaining page on the same system, copy humanised |
| 5:00 | Go-live | Contact form verified, site published |
The schedule reads like a timetable but it behaves like a relay. The sitemap agreed before the morning break governs the wireframes. The wireframe chosen before noon governs the homepage. The homepage sets the design system that every other page inherits after lunch. Fall behind in one block and the next block exposes it immediately, which is deliberate. On a conventional project this same chain exists, it is simply stretched so thin across the calendar that nobody can see it.
The first hour is the whole project in miniature
Nine to ten produces no design at all, on purpose. It converts the written brief into a page list, an address structure and a single conversion path: the one route a visitor takes from arriving to doing the thing the business needs. It also names the question each page has to answer for a searcher, because a page with no intent behind it is decoration.
I hold this hour sacred because most small business sites do not fail at colours or code. They fail here, at structure, and the failure gets papered over with visuals for months before anyone admits it. When the sitemap is honest, every later decision has something to be judged against. A vague one turns every later decision into a coin toss. A later piece in this series takes this hour apart properly.
Structure before lunch, substance after
Mid-morning is layout. The agent produces three genuinely different rough directions for the four page types a business actually needs: home, about, one content page and contact. The person in the chair judges the three against the brief and the conversion path, picks one, and has it developed to high fidelity. Options first, taste second. Skipping the options is how sites end up looking like everyone’s template.
Before lunch the chosen wireframe becomes a live homepage, and this is the block where the design system is born: one set of named values for type sizes, colours and spacing that every element refers back to. My own rebuild runs on a list of just eighteen such values, and the smallness is the discipline. Every page built after lunch inherits the list instead of inventing its own greys.
The afternoon belongs to substance. First a full article, written to earn search ranking and to be retrievable by the AI assistants that increasingly answer questions before a click happens, with metadata, structured data and internal links configured rather than left for later. Then the remaining pages are assembled on the homepage’s system, imagery is placed, and all the copy goes through a humanising pass that hunts down the tells of machine writing before any of it is published. I have written before about why that judgement layer decides whether AI output is an asset or slop.
The last hour is the one most projects never schedule. The contact form is tested end to end, meaning a real message sent and received, not a form that merely looks present. A pre-launch checklist covers indexing, SSL, mobile rendering and load speed. Then the site is published to a live address and each builder presents it against the brief they started with at nine.
What the person in the chair actually does
Not coding. At no point does anyone write code, and I would send a participant home if they tried, because the day is not a programming course. The human work is directing: stating what you want precisely, in writing, then judging what comes back critically. The rest is knowing your own business well enough to spot when something plausible is wrong. Speed comes from the agent. The brief, the taste and the veto stay with you.
If you have ever briefed an agency and could tell strong work from weak across a boardroom table, you already hold the difficult half of this skill. The unfamiliar half is smaller than people fear, and one day of supervised repetition is enough to make it stick.
The honest preconditions
I am wary of anyone selling a miracle, so here is what the day quietly depends on. The tooling is set up and verified before anyone sits down, as homework with a checklist, because environment problems are the death of any hands-on room. The brief, brand assets and raw copy material arrive written, since no agent can invent what a business actually knows about itself. And the room stays small, with a second pair of hands floating for the machine that misbehaves, because one stuck laptop should never stall eleven others.
Arrive with those pieces in place and the day ends with your site live on a working address. Arrive without them and you would spend the day watching other people build, which is why I would rather turn someone away than let that happen.
Where this comes from
The running order is not theory. It is the sequence I used when I rejected my own website and rebuilt it: positioning, structure, system, build, gates, checklist. That rebuild took two days because I was also re-answering who the site was for. The one-day format is the same relay with that question settled beforehand, as homework, where it belongs.
Next in this series: the SME owner who was quoted five figures for a five-page site, and what this way of working does to that maths. If your company has a website project stuck in exactly that place, or you want to hear when the first public run of the programme happens, start a conversation.

