There's a conversation happening right now in every developer community I'm part of. It usually starts with something like: "I used ChatGPT to build my entire app in a weekend" or "We don't need developers anymore, AI can just do it." And every time I hear it, I have the same reaction — part admiration, part concern.
Let me be honest with you: AI has made me a faster, more productive developer. I use it every single day. But the way I use it is probably very different from how someone without a development background uses it — and that difference matters enormously, especially when you're building something people are going to trust with their data and their money.
How I Actually Use AI in My Workflow
On a typical day, I might use an AI assistant to scaffold boilerplate code, draft a first pass at a Livewire component, debug a tricky SQL query, or generate a migration I'd otherwise write from memory. It's like having a very fast junior developer sitting next to me — one who never gets tired, never complains, and has read an enormous amount of code.
The key phrase there is junior developer. I still review everything. I still question it. I still know when it's confidently wrong — and it absolutely does get things wrong.
With 25 years of experience behind me, I can spot a security hole, a poorly structured query, an n+1 problem, or a race condition in a matter of seconds. AI gives me speed. My experience gives me judgement. Together, they're genuinely powerful. Apart? That's where things start to unravel.
The Real Benefits When You Know What You're Doing
For experienced developers, the benefits of AI-assisted development are real and significant:
Speed to market is the obvious one. Tasks that used to take me an hour might take twenty minutes. Boilerplate that I'd have written from scratch gets generated in seconds. That compounds across a project — we're talking meaningful reductions in build time.
Cost-effectiveness follows naturally. If I can deliver a Laravel SaaS MVP in six weeks instead of ten, that's a real saving for my clients. They get to market faster, they start learning from real users sooner, and they're not burning budget on repetitive groundwork.
Reduced cognitive load is the one people talk about less, but I find it genuinely valuable. Offloading the mechanical stuff frees up mental energy for the architectural decisions, the user experience thinking, and the business logic that actually makes a product good.
But — and I can't stress this enough — these benefits only materialise reliably when the person using the AI knows enough to validate what it produces.
Where It Goes Wrong for Less Experienced Developers
Here's the uncomfortable truth. AI makes it easier than ever to build something that looks like a finished product without actually being one.
I've been brought in to rescue a handful of projects over the past couple of years where a non-technical founder, or a very junior developer, had used AI to build an MVP. On the surface, these apps looked great. Clean UI, reasonable UX, all the expected features. But underneath?
Mass assignment vulnerabilities sitting wide open in Eloquent models. No CSRF protection on forms. Passwords being stored without hashing. API keys hardcoded directly into blade templates. Database queries that worked fine with ten users and would collapse under a hundred. No queue handling for long-running processes. Authentication flows that could be bypassed with a simple URL manipulation.
None of these things were malicious. The people who built them were working hard and doing their best. But AI doesn't know your application. It doesn't know your threat model. It doesn't know that the "users" table it's generating code for contains sensitive medical information. It generates plausible code, not necessarily correct code for your specific context.
And here's the cruel irony: the better AI gets at generating confident-looking code, the harder it becomes for someone without experience to spot the problems.
MVPs Built With AI: A Double-Edged Sword
I genuinely believe AI-assisted development is one of the best things to happen to the startup ecosystem in years. The ability to get a real, working MVP in front of users quickly — without spending £50,000 on development — is transformative.
But there's a spectrum here that founders need to understand.
At one end, you've got a simple internal tool, a landing page with a form, a lightweight admin dashboard — things with limited exposure, limited data sensitivity, limited complexity. An AI-assisted build by a capable non-developer can work reasonably well here. The risks are manageable.
At the other end, you've got a SaaS product handling user accounts, subscriptions, payments, personal data — maybe operating under GDPR. Here, the gap between "it looks like it works" and "it's actually production-ready" is enormous. And the consequences of getting it wrong aren't just technical. They're legal, financial, and reputational.
My Honest Advice
If you're a founder thinking about using AI tools to build your MVP yourself, I'm not here to tell you not to. But I would encourage you to be honest with yourself about where your product sits on that spectrum.
For a quick prototype to validate an idea? Go for it. Learn as you go. It's a brilliant time to be a curious non-developer.
For something you're going to ask real users to trust with their data? Please get an experienced developer involved, even if only for a security review before you go live. It doesn't have to be a full build engagement — even a focused code audit can catch the things that AI confidently missed.
And if you're a developer reading this — whether you're junior or mid-level — my advice is to use AI heavily, but invest equally heavily in understanding why the code it generates does what it does. The developers who are going to thrive in the next decade aren't the ones who can prompt AI the best. They're the ones who can prompt it well and know when to trust it, when to push back, and how to fill in the gaps it leaves.
The Bottom Line
AI is a remarkable tool. In experienced hands, it accelerates delivery, reduces cost, and genuinely makes building software more enjoyable. In inexperienced hands, it can produce something that looks production-ready but isn't — and the gap between those two things can cause real harm.
Experience isn't about memorising syntax or knowing every Laravel helper method. It's about judgment — the ability to look at generated code and know whether it's trustworthy. That judgment still has to come from somewhere, and right now, no AI can replace it.
Use the tools. Embrace them. Just make sure you — or someone on your team — has the experience to know what to do with what comes out the other end.