I used to think skipping design because of AI was the move.
Like a lot of people, when AI went mainstream, I convinced myself that as a designer I could bypass the whole design phase. Go straight from idea to build. Prompt it, generate something that looks like a product, wire it up, and ship. It felt efficient. It felt modern. It felt like the obvious next step in how products should be built.
It also turned out to be completely wrong.
That thinking didn’t last very long once I started building out the workflow for our ventures pipeline. Very quickly, it became clear that skipping design doesn’t remove the work, it just delays it. You still end up asking the same questions, just at the worst possible time—while you’re already building. And instead of making progress, you find yourself going in circles, tweaking things, rewriting flows, trying to understand what you should have been building in the first place.
Design, I realised, is not a step you can skip. It’s the thing that makes everything else make sense.
Most people still reduce design to UI. The screens, the colours, the layouts, the part that feels visible and tangible. That’s what gets labelled as “UI/UX,” and for many, that’s where the understanding of design begins and ends. But that’s only the surface layer. The real value of design sits underneath that, in the thinking that happens before anything is drawn or built.
Design thinking is about understanding the problem before you rush into solving it. It forces you to sit in the ambiguity a little longer and ask better questions. Who are we actually building for? What problem are they experiencing in their real lives? Why does this matter enough for them to change behaviour or pay for a solution? What should exist, and just as importantly, what should not?
Frameworks like the Design Sprint exist for a reason. They are structured ways to force clarity before execution. Over a few days, you take a messy idea and break it down into something testable. You map out the problem space, identify a specific target user, sketch different approaches, make decisions about what is worth pursuing, and then put a version of that idea in front of real people to see how they react. It is not about perfection, it is about learning quickly and intentionally. That process alone eliminates a huge amount of guesswork, and more importantly, it aligns everyone involved around what actually matters before any code is written.
AI skips all of that.
If you are just “vibe coding” your way into a product, you are essentially outsourcing the thinking. You prompt your way into interfaces and flows without fully understanding why they exist. The outputs can look convincing at first glance, but they are built on generic assumptions, not your specific context. AI does not know your user. It does not understand your constraints, your market, or the trade-offs you need to make. It gives you something that seems reasonable, but not necessarily something that is right.
So you end up in loops. You build something, realise it doesn’t quite make sense, adjust it, rebuild it, prompt again, and repeat the cycle. Instead of solving a defined problem, you are trying to discover the problem while already building the solution. That is where a lot of time gets lost.
One of the biggest things you gain from proper design thinking is a much deeper understanding of your user. Not the surface-level persona that sits in a slide deck, but a real understanding shaped by observation, conversation, and context. This is where AI has a significant blind spot. It operates on available data, and while that can be powerful, it is often incomplete.
In many environments, especially in markets like fintech across Africa, there are large gaps between what is happening in reality and what is captured in data. Cash transactions, informal systems, shared behaviours, and cultural nuance do not always translate neatly into datasets. If you rely only on what AI can see, you are working with a partial picture.
There is no real substitute for speaking to users, observing how they behave, and understanding the details of how they navigate their world. That is where the most valuable insights come from. AI becomes useful after that point, when you already have raw insight and need help making sense of it. It can help synthesise what you have learned, highlight patterns, and even suggest directions, but it cannot replace the process of discovering those insights in the first place.
When you take the time to do this properly, something shifts in how you define your product. You are no longer just assembling features, you are shaping an experience with intention. Every flow starts to have a reason. Every screen is there to move the user forward in a way that makes sense for them. You understand what needs to be solved now and what can wait. You start to see the path from a simple MVP to something more scalable and robust.
If you skip this and rely on AI-generated flows, you often end up with something that looks complete but feels disconnected. The journey might seem logical at a glance, but when you actually use it, friction appears in unexpected places. Important steps are missing, or unnecessary ones are introduced. You spend more time fixing those issues than you would have spent thinking things through at the beginning.
This is why, as outdated as it might sound, starting with low fidelity still matters. Sketching ideas out, mapping flows, and intentionally designing each step forces clarity. It slows you down just enough to think properly. It also gives you control. Instead of inheriting decisions from a generated output, you are making them yourself, based on what you know about your users and what you are trying to achieve.
That clarity carries through into the details. And the details are where things start to feel different.
When you understand your users properly, you begin to care about things that go beyond layout and visuals. You think about tone, language, and how the product communicates. You choose colours and styles that resonate with your audience, not just something that looks trendy or “AI-generated.” You create something that fits your customer, rather than something that could apply to anyone.
I experienced this firsthand recently when I bought a third-party Xbox controller from a local South African site I had never heard of before. The interface itself was not particularly impressive, but the experience stood out. The site used local slang in a way that felt natural and relevant. The products were curated rather than overwhelming. The messaging was consistent from browsing to checkout to delivery. It felt intentional. I could picture the person behind it and understand who they were speaking to.
They likely used a standard platform, probably something like Shopify. There was nothing technically groundbreaking about it. But they understood their audience and paid attention to the small details that made the experience feel right. The UI was not perfect, but the overall experience was lekker. And that is what made me trust them enough to buy, and want to come back again.
This is where design thinking extends beyond the product itself and into the brand. When you understand your users deeply, you are able to build something that people can relate to. You know how to speak to them, where to find them, and what matters to them outside of your product. That shapes everything from your messaging to your marketing to how your product evolves over time.
A simple example is something like a budgeting app. If it is built in a generic way, it fades into the background. It becomes one of many similar tools, competing on features alone. But if it is designed for a specific audience, such as young creative business owners, the entire approach changes. The features become more relevant, the language becomes more relatable, and the product starts to feel like it was made for someone, not just anyone.
At that point, you are no longer just building a tool. You are building something that connects.
There is a lot more value in design thinking than people are giving it credit for right now. In a world where it is easier than ever to generate something that looks like a product, the real advantage comes from understanding what should be built and why. That is not something you can shortcut.
We have started bringing this thinking back into our own workflow in a much more deliberate way. Slowing down at the beginning, using structured approaches to unpack the problem, spending more time understanding our users, and then using AI where it actually adds value. Not to replace the thinking, but to support it. To move faster once direction is clear, to explore ideas more broadly, and to execute with more efficiency.
If you are a founder and you want to follow along, learn some of the methods we are using, and see how we are applying this in practice, subscribe to the Dive Into Product blog.
We are sharing the process, not just the outcomes.