AI and digital strategy
FamilySafe had a working product and no category. I defined one, built the brand around it, wrote the content strategy and specified the site, on my own, using AI wherever it earned its place. Strategy to a staged site took 10 weeks, with 52 help centre articles planned and structured, 30 brand images produced from a written brief, and 3 WCAG 2.2 AA violations fixed at design token level.
FamilySafe helps families get their affairs in order before anyone needs them to be. Where the documents are. Who can see what. What happens when someone dies and a relative is suddenly dealing with six banks, a mortgage provider and a probate process nobody explained to them.
The product worked. The founder knew what he had built. The problem sat above all of it, in the words.
Nobody could say in one sentence what FamilySafe was. The working category at the time was Preparedness Infrastructure, which is the sort of phrase that sounds substantial in a board pack and means nothing to a fifty three year old woman searching for help with her mother’s accounts at eleven at night. Underneath that, the pricing was already well developed and the partner programme was genuinely sophisticated. Neither was doing any work, because the framing above them was vague.
Three months. No agency budget. One person.
Category before anything else. I replaced Preparedness Infrastructure with Family Readiness for consumers, and Client Readiness for the partner facing version that advisers, solicitors, brokers and care providers see. The consumer tagline moved from “Your life, organised. Your family, protected.” to “Your life, organised. Your family, ready.” Protected is passive and slightly frightening. Ready is something you can be proud of being.
Then the rule I am most pleased with, written into section 3.0 of the website brief so it could not quietly erode later.
The two layer rule. Industry vocabulary never appears in consumer copy. A grieving daughter does not want to read about readiness thresholds. She wants to know which form the bank needs.
It sounds like a small editorial policy. It is actually the whole brand. It governs the help centre, the article library, the LinkedIn posts and every page on the site, and it is enforced document by document rather than left to good intentions.
This is the part no tool could have done. It came out of reading the competitor set, the live site and the market, then making a judgement about where the value actually sits.
Strategy and positioning
The category claim, the partner variant, the brand promise, the five pillars, the revised consumer tagline, and the two layer rule codified with a written rationale so future contributors understand why rather than just what.
Website specification
Brief v3.2, the definitive version, written as an execution document for Claude Design and Claude Code rather than as a strategy deck that someone else would have to translate. Structure, page by page intent, article threading, search strategy.
Brand imagery
Thirty photo realistic images produced from a single written style brief: eighteen for the consumer family pages, twelve for the professional sector pages. A master style block held the look constant so the set reads as one commissioned series rather than thirty unrelated stock photographs. Every image shipped with a filename and an alt tag written at the same time, not retrofitted.
Content architecture
A help centre plan of fifty two articles across six topic groups, one row per article, priority tiered so it could be approved, cut or added to before anyone wrote a word. Alongside it, a twelve month content plan covering the article library and the social arc, with the five pillars running through both.
Founder communications
Forty LinkedIn posts written in the founder’s own voice, mapped across ninety eight slots over ten weeks so the same post never lands near itself. Thirty speak to families in everyday language. Ten speak to partner firms in the Family Readiness vocabulary. They never mix inside a single post, because that is the two layer rule again.
Accessibility
An axe-core audit against the preview site found three WCAG 2.2 AA violations: hero photo toggles below the 24 by 24 pixel target size, clay accent text at 3.49:1 contrast, and amber status text on the probate advisor at 4.39:1. All three were fixed at the design token and brand guide level rather than page by page, so the same problems cannot reappear the next time someone builds a component, then verified with a re-scan. A publishable accessibility statement was drafted alongside it.
I want to be specific here, because vague claims about AI are exactly the thing I spend my working life arguing against.
Research at a scale the budget could not otherwise buy
Competitor analysis, market review and a full read of the existing site, compressed from weeks into days.
Production capacity
Thirty brand images from a written brief instead of commissioning a photographer, a location and a day. Content drafted against a locked voice at a volume that would normally need a small team. This is the difference between having a plan and having something you can actually put live.
Build
The website brief and the accessibility fix brief were both written as instructions for Claude Design and Claude Code. I wrote the specification. The machine did the typing, and did it faster and more consistently than a junior developer would have.
Inside the product
Readiness Intelligence, the AI capability in FamilySafe itself, went from an unclear feature name to a defined part of the proposition with a place in the roadmap.
The category call. The two layer rule. The decision to fix accessibility at the token layer instead of the page layer, which cost more thinking up front and saves every future build. The editorial policy for publishing about death, money and probate, where Google’s standards for this kind of content are unforgiving and getting it wrong does real harm to real people. The judgement about what to build first when everything is urgent.
AI gave me the output of a team. It did not give me the decisions.
That distinction is the whole argument. Point a language model at a muddled strategy and it will produce muddle at volume, beautifully formatted, in about nine seconds. Point it at a clear one and you get a quarter’s work in ten weeks. The multiplier is real. What it multiplies is whatever judgement you brought with you.
Strategy, brand system, content architecture and the accessibility work are complete. The site is in staging at preview. The help centre articles and the twelve month library are structured and ready to write. Social switches on in August.
I should be straight about what this case study is and is not. It is a record of how much can be built, by one person, in a short window, using AI properly. It is not a record of revenue or conversion, because the site has not been live long enough to produce honest numbers. Those follow in a second version after the first full quarter of trading.
Anyone showing you conversion uplift figures ten weeks into a rebuild is showing you noise.
Because the framing above the product was what was not working. The pricing was already well developed and the partner programme was genuinely sophisticated, and neither was doing any work while nobody could say in one sentence what FamilySafe was.
It earned its place on volume work: 52 help centre articles planned and structured, 30 brand images produced from a written brief. The category decision, the words that carry the positioning and the judgement about what to cut were not AI work, which is the point.
Deciding the category first, so every later decision had something to test against. Most of what makes a project slow is re-litigating the framing, and that had been settled before the content strategy or the site specification started.
Strategy work of this shape is digital strategy. On using AI for the volume work while keeping the judgement human, see the benefits of AI-generated content with human review.
Most clients begin with a conversation. No pitch, no hard sell.
Just a straightforward discussion about where you are and whether I can help.
Book a free 30-minute call