I’m Ricardo, and I run Vertex Platform Solutions as a solo developer. When I tell people I can deliver a professional business website in under a week, they usually assume I’m using templates or cutting corners. The truth is more interesting: I use AI tools extensively, but not in the way most people think.
This isn’t a hype piece about AI replacing developers. I’m still writing code every day. But my workflow has changed dramatically, and I want to share what actually works.
The Tools I Actually Use
I’m not going to pretend I’ve found some magic AI that builds websites for me. Here’s my honest stack:
Claude (Anthropic) - My primary coding assistant. I use it for generating component structures, debugging obscure errors, and writing repetitive code. It’s particularly good at Astro and React patterns.
GitHub Copilot - Autocomplete on steroids. Best for finishing lines of code I’ve started, suggesting standard patterns, and catching syntax errors before I run the code.
ChatGPT - Strategic planning and content structure. I use it more for thinking through architecture decisions than writing code.
Midjourney - Conceptual design work and client presentations. Not for final assets, but for exploring visual directions quickly.
What a Week Actually Looks Like
Let me walk through a real project from last month: a 6-page website for a Manchester-based HVAC company. If you’re looking for a professional website without the agency markup, check out our business website service.
Day 1: Discovery and Architecture (AI Contribution: 30%)
I spent the morning on a video call with the client, understanding their business, competitors, and goals. AI can’t do this part - you need human judgment to understand what problems you’re actually solving.
After the call, I used Claude to help structure my notes and identify technical requirements. I asked it to suggest suitable tech stacks based on their needs (static site, SEO focus, easy content updates). It suggested Astro with a headless CMS, which aligned with my thinking.
I designed the information architecture and sitemap myself. AI can suggest structures, but it doesn’t understand the client’s specific context well enough to make these decisions.
Day 2-3: Design and Component Building (AI Contribution: 65%)
This is where AI shines. I created the design system in Figma myself - colour palette, typography, spacing tokens. But once I had that, I used Claude to help me build the component library.
I’d describe what I needed: “Create an Astro component for a service card with an icon, title, description, and optional CTA link. Use these design tokens…” and paste my CSS variables. It would generate clean, accessible code that matched my design system.
I still reviewed every line. AI occasionally makes assumptions about accessibility or responsive behaviour that aren’t quite right. But instead of writing 200 lines of boilerplate myself, I was reviewing and refining 200 lines generated in seconds.
Day 4: Content and SEO (AI Contribution: 40%)
I wrote the hero headlines and main value propositions myself. This requires understanding the client’s voice and competitive positioning - AI tends to produce generic marketing speak.
But for service descriptions, FAQs, and technical content, I used ChatGPT to expand bullet points into proper copy. I’d write the key points, and AI would flesh them out. Then I’d edit for tone and accuracy.
For SEO metadata, I used Claude to generate variations of titles and descriptions based on target keywords. It’s faster than manually writing 20 variations, and you can pick the best ones.
Day 5-6: Integration and Testing (AI Contribution: 50%)
This phase involves a lot of debugging and browser testing. AI is brilliant for fixing obscure CSS issues or JavaScript errors. I’d paste the error message and relevant code, and Claude would usually spot the problem immediately.
I had an issue where the contact form wasn’t submitting correctly on iOS Safari. After 20 minutes of manual debugging, I asked Claude to review the code. It spotted that I was using a form attribute that Safari handles differently. Fixed in two minutes.
For cross-browser testing, I still do the manual work. But when I find issues, AI helps me fix them much faster.
What AI Is Actually Good At
After 18 months of working this way, here’s what I’ve learned AI excels at:
Boilerplate and repetitive code - If you’re writing similar components with slight variations, AI can generate them instantly. Form inputs, card layouts, list items - all perfect use cases.
Debugging syntax errors - Paste your error message and code, get a fix in seconds. This alone saves me hours every week.
Explaining unfamiliar code - When I’m working with a library I don’t know well, AI can explain what specific functions do and suggest implementations.
Accessibility improvements - AI is quite good at suggesting ARIA labels, keyboard navigation, and semantic HTML improvements.
Writing tests - I describe what behaviour I want to test, and AI generates the test structure. I still review and refine, but it’s much faster than starting from scratch.
What AI Is Terrible At
Let’s be honest about the limitations:
Understanding business context - AI doesn’t know why your client needs a website, who their customers are, or what makes them different from competitors. You need human judgment for strategy.
Design taste - AI can generate layouts, but it has no aesthetic judgment. Every AI-generated design looks similar - generic, safe, forgettable. You need a human to make design decisions that reflect the brand.
Complex business logic - If you’re building custom e-commerce workflows, booking systems, or integrations, AI will give you a starting point but you’ll spend significant time refining and fixing edge cases.
Security considerations - AI might use outdated security practices or miss important vulnerabilities. You can’t trust it to handle authentication, payment processing, or data protection without careful review.
Project management - AI won’t tell you when you’re over-complicating things, when deadlines are unrealistic, or when the client needs to make a decision. You’re still running the project.
The Economics of Solo + AI
Here’s why this model works for service businesses:
A traditional agency charges £5,000-£12,000 for a 5-page business website and takes 6-8 weeks. They have overhead: office space, multiple salaries, project management layers.
I charge £3,000-£6,000 and deliver in under two weeks. My overhead is minimal - just software subscriptions (about £80/month for AI tools) and my time.
The client gets:
- Faster delivery
- Direct communication with the person building their site
- Professional quality without agency markup
- Someone who actually cares about the outcome
I get to work on projects I find interesting, charge sustainable rates, and maintain a good quality of life. No 60-hour weeks, no office politics, no pitching to clients I don’t want to work with.
The Skills That Still Matter
If you’re thinking about working this way, you absolutely need:
Strong fundamentals - You must understand HTML, CSS, JavaScript, and web performance. AI makes you faster if you know what you’re doing; it makes you dangerous if you don’t.
Design judgment - Someone needs to make the aesthetic decisions. AI can’t tell you if something looks good or matches the brand.
Client communication - Understanding what clients actually need (versus what they say they want) is still entirely human work.
Quality control - You’re reviewing everything AI produces. If you don’t know what good code looks like, you’ll ship mediocre work.
Problem-solving - When something breaks in production, you need to diagnose and fix it. AI can help, but you’re driving the process.
What This Means for Clients
If you’re a business owner considering hiring a solo developer who uses AI, here’s what you should know:
It’s not a red flag - Using AI tools is like using any other productivity software. You wouldn’t penalise a designer for using Figma instead of drawing by hand.
But ask questions - How much code do they write versus generate? Can they explain every part of their codebase? Do they understand security and performance?
Focus on outcomes - Does the site work? Is it fast? Is the code maintainable? Is it accessible? These matter more than how the code was written.
Faster doesn’t mean worse - I’m not cutting corners; I’m eliminating busywork. The strategy, design, and quality control still take time and expertise.
The best question you can ask: “Can you explain why you made this technical decision?” If they can’t, that’s a red flag - whether they used AI or not.
The Honest Limitations
I’m not building everything with this workflow. Complex web applications with custom backend logic? That’s a different project entirely. E-commerce sites with hundreds of products? I’d partner with specialists or recommend a platform like Shopify.
This approach works well for:
- Business websites (5-15 pages) - see our business website service
- Marketing sites and landing pages
- Portfolio and showcase sites
- Content-focused sites with blogs
- Service business websites
It’s less suitable for:
- Complex web applications
- Custom e-commerce platforms
- Sites requiring extensive backend development
- Projects with highly specific technical requirements
What’s Next
AI tools are improving rapidly. Claude released a new model last month that’s noticeably better at understanding design systems. GitHub Copilot is getting better at suggesting entire functions instead of just completing lines.
But I don’t think solo developers will become obsolete or that AI will replace agencies. The core skills - understanding client needs, making design decisions, architecting scalable solutions - are still human work.
What’s changing is that talented individual developers can now compete with larger teams on delivery speed and quality. That’s good news for clients who want direct access to skilled people without agency overhead.
If you’re a developer, my advice is simple: learn the fundamentals deeply, then use AI to eliminate the tedious parts. Don’t use it to skip understanding; use it to spend more time on the work that actually matters.
And if you’re a business owner looking for a website, ask about outcomes and process, not whether someone uses AI tools. The proof is in the final product, not how it was made.
I still write code every day. I still debug frustrating issues and refactor messy implementations. But I spend more time on strategy, design, and client communication - the parts of the job I actually enjoy. That’s the real benefit of AI: not replacing developers, but letting us focus on being better at what we do.
