Built with AI
How this site was built
About sixty hours of work, over ten days. I set the direction, defined the audience and shaped what you see here. AI helped me develop the copy and handled the coding and technical checks. The work started with a clear brief: who the site was for, what it needed to do, and the standards it had to meet.
- My experienceTwenty years of my own work, assembled and checked
- The briefThe audience, purpose, design and requirements
- Build and refineSuccessive versions, shaped by feedback
- Review and approveTechnical checks and my approval on every page
Approval
Brief · build · review · refine
Who did what
I set the direction. AI helped me build it.
I defined who the site was for, the commercial strategy and the services it would offer. I shaped the structure, the brand, the visual design and the voice. The substance came from my own work, and I decided what was relevant to the people I wanted to reach.
AI was a sounding board throughout. We developed copy, explored ways to explain the ideas and checked that the pages worked together. It helped connect the source material, check links and spot duplication. I challenged its suggestions and made the final decisions.
AI handled the coding, engineering tools and automated tests. I reviewed the result as a reader and a customer would: whether it made sense, looked right and worked properly. Every page needed my approval before publication.
The work in numbers
~60
hours spent developing the site, over ten days
11
planning documents before coding began; 13 across the build
48
written requests guiding changes to the site
182
issues and decisions recorded as the work progressed
49
sets of automated tests, covering two browsers, accessibility and performance
1
final approver: me
As at . The hours are estimated from the times when versions of the site were saved, excluding gaps of more than ninety minutes. The other figures count the planning documents, written change requests, recorded issues and automated test files kept with the project.
Rules of engagement
A clear brief before coding began.
Client information stays private.
Confidential client material stays out of the website project, including its working notes, code, tests and screenshots. The site draws on what I can share about my experience. A client’s private information remains theirs.
I approve every page before it goes live.
A page cannot be included in the public site until I have approved it. If its copy or appearance changes, it comes back for review. The software enforces that requirement.
I own the facts about my work.
I decide what I can say about my own experience and take responsibility for it. Facts about other organisations or published research carry a source. AI can help express the account; it cannot invent the experience behind it.
Build and review are separate jobs.
I use separate AI sessions to develop the instructions, build the site and review the result. A review produces suggestions and questions for me to resolve. Neither writing the code nor checking it gives AI the authority to approve a page or publish it on its own.
The writing has to sound like me.
I defined the audience, tone and language before drafting the pages. AI suggests wording and checks for repetition and phrases I do not want. I read the result and decide whether it sounds natural, says what I mean and earns its place.