A year ago, if you'd asked me to build a customer-facing portal, I would have told you we'd need to hire multiple full-time frontend engineers. I come from a backend development background;I hadn't really done much frontend work. We'd built some internal tools but internal users are forgiving. External users? That's a completely different standard.
A year ago, this timeline was impossible. Today, it's just what AI-assisted development makes achievable when you use it strategically.
When we decided to build the Newstex Customer Portal,a web-based dashboard where our customers can discover, manage, and license content from thousands of publishers,I knew we had a challenge. We needed a modern, professional interface that could handle sensitive customer data securely. We needed it fast. And we didn't have a frontend team.
What changed? AI coding tools like Cursor that excel at helping build user interfaces.
Finding the balance
The challenge wasn't just about building a UI.t was about maintaining security while using AI assistance. Coming from the backend world, I'm hardwired to think about security first. You always assume your frontend is not a trusted client. All real security has to happen on the backend.
Use AI where it amplifies your capabilities. Use human expertise where it matters most.
When you're exposing systems to external users, you double down on that principle. Every customer who logs in should see exactly what they need to access their own data and nothing more.
This security-first mindset actually made AI-assisted development possible. If your backend is rock-solid and hand-coded, you can be more aggressive with AI on the frontend. Even if the vibe-coded frontend has a bug, worst case is something breaks visually.
This architectural decision, keeping the backend hand-coded while using AI for the non-critical frontend, let us move incredibly fast without compromising what matters most.
Managing AI like a team
What's really cool about tools like Cursor is how they make UI development feel like working with a team. I can give it our API specifications and a design concept, then iterate conversationally. "Increase the padding around the user circle." It just does it.
A small company like ours can achieve what would normally require a much larger team.
Right now, I typically work on three to five features simultaneously. Each has one or two AI agents running so I might be managing five to ten agents at once. It genuinely feels like I'm approving pull requests from different developers except I'm also the only engineer. What's even better is by using pull requests in Github, I can even have AI Agents review code. Not only do I have AI Developers, I have AI Agents helping with quality assurance.
That's the force multiplier effect. A small company like ours can achieve what would normally require a much larger team.
The timeline transformation
We went from concept in February to production launch in October–under a year. Without AI assistance, this would have been challenging with our resources. We would have needed multiple dedicated frontend engineers, and even then, the timeline would have stretched significantly (Onboarding a new development team would have taken 3-6 months at a minimum).
It genuinely feels like I'm approving pull requests from different developers except I'm also the only engineer.
We made deliberate tool choices too. I personally tested multiple different frontend frameworks, including SolidJS, Angular, and even HTMX. Even though I personally preferred using Solid since it seemed more familiar to me, we picked React specifically because AI coding assistants work much better with it. The ecosystem of examples and patterns makes these tools more effective.
When customers notice
Here's what made it worthwhile: A customer told us this is the best thing they've seen in ten or twenty years in this business.
As a backend developer, I'm used to feedback like "no problems with the system." That's the win—invisible infrastructure that just works. Being told that customers not only see what we built but love using it? That's huge.
When I look at the portal daily, I see things I'd still improve. But sometimes I forget to step back and realize it's actually a pretty good system and people genuinely like it.
We went from concept in February to production launch in October—under a year. Without AI assistance, this would have been challenging with our resources.
Strategic placement is everything
The key lesson isn't "use AI for everything." It's knowing where to use it and where not to. AI excels at generating UI code when you can give clear specifications and iterate quickly. It's less suitable for security-critical backend logic where you need complete control.
This isn't about AI replacing developers. It's about strategic force multiplication. Use AI where it amplifies your capabilities. Use human expertise where it matters most. And design your architecture so that even if AI-assisted parts fail, they fail safely and securely.
The security split between frontend and backend made everything else possible. It let us move fast without compromising on what we couldn't afford to get wrong.
A year ago, this timeline was impossible. Today, it's just what AI-assisted development makes achievable when you use it strategically.
About the author,
Christopher Moyer
