How this site was built.
Delivery decisions, review controls and the limits of the verification completed so far.
The tool release gates
The Finder is implemented and release-gated. It needs 30 to 40 approved use cases, scored examples, malicious-input checks, tested timeouts, rate and spend limits, an off switch and an accurate retention notice.
Phil must read and approve the scored outputs before release. A fallback to static department examples requires his explicit approval and a cut-off date agreed in advance. No release or fallback has been approved.
The policy selector needs written, versioned clauses with review dates and reviewer names. Every distinct document path must be generated and read before release.
A fallback to a static template requires the same date and approval conditions. No template has been approved for download.
The generative policy builder stays deferred. Nothing on this page claims either tool has been released.
Decisions
- Two fonts, hosted on the site.
Playfair Display for headings and DM Sans for everything else. The brief asks for the font files to be served from this site so visitors do not make a request to Google.
- Text sizes.
Body text is 16px or larger and labels are 14px or larger on the pages rebuilt for this version.
- Colour roles come from calculated contrast ratios.
Coral is used for large numerals only. Mint is never used as text on paper. Ink, teal and white pairs are used for text and buttons.
- Examples are text, not a chatbot.
The department examples on the home page are written out in the page. No AI tool is live on this site, and nothing on it pretends to be one.
- One label per action.
Consulting is always Book a call and goes to the contact page. Speaking is always Check availability and goes to the speaking enquiry form.
- Navigation by keyboard.
Menus open with the keyboard, Escape closes them and returns focus to the button that opened them. The mobile menu is a full-height panel.
- Short motion, with a plain fallback.
Page transitions, a scroll story, quote reveals and one count-up use support checks. Reveals finish within five seconds. Reduced motion and unsupported browsers get the plain page. No video plays automatically.
- The AI Opportunity Finder is built but not released.
It uses fixed choices, shows an AI label and processing disclosure before submission, and has owner controls for the library, scored examples and release. It stays unavailable to visitors until the owner approves its release.
- Insights stay hidden from the header until there are three published articles.
A menu item that leads to an empty page is a dead end.
- Only real reviews appear.
The results page and quote sections show reviews a client has approved for publication, and nothing else.
Verification limits
This site has not been certified against any accessibility standard. It has not been tested on a range of real phones and assistive technology. Page speed targets have not been measured in the field.
I will add results here when the checks are done, with what was tested and what was found.
AI and data policy selector (coverage and availability shown honestly)