Website rebuilds have a reputation. Big teams, long timelines, significant budgets – and results that often disappoint. Our previous site took a team of seven and six months to deliver.
So when it came time to revisit the site, we started with a modest goal: make some improvements to what we had. We began prototyping with AI early, and it became clear quickly that starting from scratch would actually be faster, and the quality ceiling significantly higher, than trying to improve what existed. The scope changed. The approach didn't.
The result was a completely rebuilt website, delivered in two months, by two people – Jon Burton, our Growth Marketing Manager, and Ben Dallimore, our Solutions Consultant, who led development and AI implementation – both working alongside their regular responsibilities.
The conditions that made it possible
This wasn't simply a story of AI doing the heavy lifting. The technology mattered, but it wasn't the deciding factor. Two things were.
Domain knowledge
This is the constraint that quietly undermines most website projects. When the people doing the work – writing copy, making content decisions, shaping the information architecture – don't have deep familiarity with the product and the market it serves, that gap has to be bridged somehow. It rarely is, fully. The result is content that sounds plausible but doesn't quite connect with the people it's meant to reach.
Affinity Payroll offers specialist products and services in a genuinely complex domain. Australian and New Zealand payroll involves layered compliance obligations, nuanced award interpretation, and real consequences when things go wrong. Our customers – payroll managers, HR managers, finance managers – know this territory intimately. Speaking to them effectively requires the same level of familiarity.
On this project, that knowledge was inside the team doing the work, not something we were trying to transfer to people outside it.
Strategic clarity
Before any design or development began, we worked through the StoryBrand framework to get precise about who we're speaking to, what problem we're solving, and what Affinity's role is in the customer's story. Positioning and messaging were locked in before a single line of code was written.
That foundation kept the entire project coherent – fewer revisions, less ambiguity, faster decisions at every stage.
Clarity before code – and a new kind of project foundation
That strategic work didn't just inform the build. It became the build.
All of the documentation – brand guidelines, StoryBrand scripts, SEO and AEO strategy, metadata registers, product library, tone of voice – was developed and refined in Claude.ai, organised as a project with persistent context throughout the entire engagement. Claude.ai wasn't a writing assistant bolted on to the process. It was the environment where strategy was developed, tested, and documented. Everything that needed to be true about how Affinity presents itself lived in one place, accessible and consistent across every task and every decision.
Documentation developed in Claude.ai included:
This is what made the AI-first approach genuinely different from simply using AI as a tool. Claude.ai wasn't starting from scratch on each task. It was working from a shared, documented understanding of who Affinity is, who we serve, and what we're trying to say – the same understanding the human team had developed and could interrogate and refine over time.
From documentation to deployment
With that foundation in place, Claude Code built the site – working from the documented context already established in Claude.ai. The development workflow was deliberately simple.
The deployment workflow
The distinction between the two tools mattered. Claude.ai held the strategy and context. Claude Code executed against it. That separation – thinking first, building second, with AI supporting both – is what kept the project coherent and fast.
But the important nuance remains: AI accelerated the execution of decisions that had already been made by people with the context to make them. It didn't define the positioning, determine the content priorities, or make judgment calls about what would resonate with our audience. Those things stayed human. The leverage came from keeping strategy human, building shared context deliberately, and letting AI carry the implementation load.
A word on scope
It's worth being clear about what kind of project this was. The new Affinity site is a static site – no complex functionality, no authentication, no databases. That makes it a genuinely approachable starting point for an AI-first build, and we'd encourage anyone considering a similar approach to start with a project of comparable scope.
The principles scale, but the complexity of execution does too.
What it cost – and what it costs to run
One of the less obvious benefits of this approach is the cost structure. The entire stack – Claude.ai, Claude Code, GitHub, and Vercel – runs at a fraction of what the previous project cost to build, and a fraction of what traditional hosting and maintenance arrangements typically cost to sustain.
The team that built the site is the same team maintaining and extending it, using the same tools and the same documented context. There's no agency retainer, no handoff risk, no institutional knowledge sitting outside the business.
The full stack:
What the site delivered
The new site launched on Next.js via Vercel with 90 redirects in place, a complete sitemap, structured data schema across key pages, and a metadata register governing every URL. The technical and SEO foundations are built in from the start – not retrofitted after the fact.
Page performance is significantly improved. And the governance is coherent enough that the same two people who built it can continue to develop it.
Technical foundations built in from launch:
The broader point
The conditions for a successful website project aren't primarily about team size or budget. They're about clarity – knowing exactly what you're building and who you're building it for – domain knowledge that stays close to the work, and AI tooling used to amplify strategy rather than substitute for it.
When those conditions exist, a small team can achieve things that would previously have required far greater resources. That's the real change AI has enabled. Not the replacement of expertise – the amplification of it.