.png)
The AI hype train is moving at full speed, promising to revolutionise everything from writing emails to writing code. But for full-stack developers in the trenches, the real question is: how does this translate to our daily work? Specifically, when it comes to user interface (UI) design and development, can AI move beyond being a novelty and become a genuinely useful tool?
At Optis, we encourage exploration. To get a balanced view, we sat down with two different perspectives: our full-stack developer Kenneth, and our managing partner Keshia. Their combined experience offers a refreshingly pragmatic reality check that separates the hype from the helpful.
The first step for many developers into AI is a chat interface like ChatGPT. While useful, it creates a disconnect between your code and the AI's context. This is where AI-native IDEs like Cursor.ai change the game.
"The biggest advantage of a tool like Cursor is context," Kenneth explains. "It can see multiple files in my project, which makes asking for help much more efficient than copying and pasting code into a separate chat window."
For day-to-day development, Kenneth uses AI not as a lead architect, but as an incredibly smart assistant. "Professionally, I use it for very specific tasks where I already know the answer. It’s a race of who can type it faster, me or the AI. It's perfect for boilerplate code or small, repetitive snippets."
This smart autocomplete, which can suggest entire blocks of code based on the surrounding context, is where AI currently delivers the most consistent value. It’s not about offloading thinking; it’s about accelerating execution.
So, can you prompt an AI to create a beautiful, functional UI from scratch? This is where the user's background makes all the difference.
For Kenneth, it's about quick tweaks on side projects. "I'll ask it to 'make this a bit prettier' and it does a decent job generating Tailwind CSS classes. It gets me something that looks good enough without much effort."
But what about someone who isn't coding daily? Keshia, who last coded professionally five years ago, put this to the test during a company hackathon using v0.dev, a generative UI tool from Vercel.
"My first attempt was a mega prompt explaining the entire application," she recalls. "The tool was completely lost." The breakthrough came from learning to work with the AI. "You have to feed it step-by-step. Start with 'I want a web application to manage a budget,' and then add layers. It’s a learning curve."
Crucially, her team didn't use the generated code directly. Instead, Keshia acted as a rapid prototyper. "I used it to get the designs. I could feed it our logo and say, 'use these brand colours,' or 'make it look calmer.' Then, I’d take screenshots, share them on Slack, and the team would implement them."
This "air-gapped" workflow is a powerful, low-risk way to use generative AI: as an inspiration engine and mock-up tool, not a production-code factory.
One of the strongest warnings, shared by both Keshia and Kenneth, is about the danger of over-reliance on AI, especially for those still learning the craft.
"We see it with interns," Kenneth notes. "They trust the AI so implicitly that they stop thinking critically or even opening documentation. They don't know why the code works, or even if it's correct. This often leads to solutions that are overly complex or based on outdated libraries because the AI's knowledge is frozen in time."
This is a critical point. An LLM can't tell you that a new, more efficient SDK has been released since it was trained. A senior developer can spot this instantly, but a junior developer might spend days fighting with a solution that a quick look at the docs would have solved.
The key to successfully using AI is to know what you're doing. You need the expertise to validate the AI's output, catch its mistakes, and steer it in the right direction.
Right now, AI in development is like a toddler you take along. It can handle simple helper tasks, but it can't take over your job. You still need to be the expert in the room.
Beyond correctness, there's the issue of determinism. Kenneth recalls trying to build a personal app to categorise bank statements using AI. "It was a nightmare. Sometimes it worked, sometimes it completely missed the mark. As a programmer, not knowing if your code will reliably do its job is a terrible feeling."
This is why he's cautious about using AI to generate quick prototypes for clients. "A designer can show a clickable wireframe, and everyone knows it's a mock-up. But if a developer shows a 'working' UI generated by AI, the client might think 90% of the work is done. They don't see the messy, unreliable code underneath."
For Kenneth, it comes down to accountability. "When I deliver code to a client, I want to stand behind every single line. I need to know it was thoughtfully constructed, not just dropped into my project by an algorithm."
At its core, the quality of an AI's output is a direct reflection of the user's input. The tools are powerful, but as Keshia found, a vague prompt like "make it prettier" yields generic results. A professional designer using the same tool would provide far more specific instructions about alignment, spacing, and hierarchy, achieving a superior result.
AI tools are not equipped with good taste or a deep understanding of user experience. They can assemble components, but they can't yet grasp the why behind a great design. As Keshia notes, the designs from v0 are functional, but "dry" and "mainstream". Perfect for an internal tool, but not something that will make a brand stand out. For truly exceptional and demanding projects, the creativity and expertise of a professional designer remain essential.
Ready to experiment responsibly? Here are some of the tools we’ve explored:
Whether you're a hands-on developer speeding up your workflow or a manager prototyping a new idea, the principle is the same: AI is a tool to augment human expertise, not replace it. It can accelerate, inspire, and automate, but it cannot yet own the creative or critical thinking process.
Your expertise is what turns an AI's generic suggestion into a robust, maintainable, and beautiful solution. The real value isn't in letting the AI take over, but in learning how to use it to become more efficient and creative.
