The Day the App Stopped Talking — and What I Learned About Listening to Users
When the AI-powered planning assistant I built failed to understand a user's request, I realized that no matter how smart the model, it's the people who matter most. This mistake reshaped how I approach building tools for real users.
"The Day the App Stopped Talking — and What I Learned About Listening to Users When the AI-powered planning assistant I built failed to understand a user's request, I realized that no matter how smart the model, it's the people who matter most. This mistake reshaped how I approach building tools for real users. There was a moment when Planner, my AI-powered planning assistant, stopped making sense. I had built it with the intention of helping users break down complex goals into manageable steps. It used AI to reason through dependencies, suggest next actions, and even ask clarifying questions. But one day, a user sent a request that the AI didn’t understand. It didn’t ask for clarification, didn’t offer a suggestion — it simply said, 'I don’t know what to do next.' That was the moment I realized my mistake: I had assumed the AI could understand everything, when in reality, it needed more context from the user.
I had spent weeks refining the model, training it on a variety of scenarios and planning tasks. I had assumed that if I built a good enough AI, it would handle everything. But in reality, it was the user who had to be at the center of the experience. The AI wasn’t the hero — the user was. This mistake taught me that no matter how advanced the AI, it's the human side of the equation that truly matters.
The problem wasn't with the AI itself — it was with the way I had designed the user experience around it. I had focused too much on the AI's capabilities and not enough on the user's needs. I had built a tool that could think, but not one that could listen. So I went back to the drawing board. I added more prompts that encouraged users to provide context, I made the AI ask more questions, and I designed the interface to be more conversational rather than transactional. The change was subtle, but the impact was real.
This experience wasn’t unique to Planner. It echoed across my other projects as well. When I built Arise & Shine Transporters, I had assumed that GPS telemetry and distance-based pricing would be enough to solve the logistics challenges of the transport industry. But I quickly learned that the real issue wasn’t just tracking vehicles — it was understanding the people who used them. That mistake led me to build a more user-focused interface and to add features like real-time updates and mobile notifications.
With Mwalimu Cosmetics, I had assumed that AI-assisted e-commerce would be enough to attract customers. But the mistake came when I ignored the feedback of local users who wanted more customization and easier navigation. It taught me that AI is a tool, not a solution on its own. It has to be guided by the people who use it.
Building AI-powered products for everyday life in Kenya isn’t just about building smart tools. It’s about building tools that understand the people who use them. It’s about listening more and assuming less. It’s about recognizing that the AI can’t solve everything — but the people who use it can.
This mistake taught me that the hardest part of building AI tools isn’t the code. It’s the people. It’s the conversations. It’s the feedback. It’s the willingness to listen, to change, and to grow. And that’s something I carry with me in every project I build — from Planner to Local
