The Chatbot That Said 'No' — And How I Found the Right Question

Every project starts with an assumption — and not all of them are right. When I first set out to build the chatbot for Arise & Shine Transporters, I thought it would be straightforward. I assumed that if I trained the AI on the most common customer questions, it would handle the rest. That was my mistake.

The chatbot was designed to answer questions about delivery schedules, pricing, and contact details. I trained it using the most frequently asked questions from the business’s customer support logs. For a while, it worked — until it didn’t. A user asked, 'How long does it take to deliver to Nairobi?' The chatbot responded with a generic answer about delivery times, not the specific details of the request.

That moment was a wake-up call. I had assumed that training on the most common questions would cover all scenarios, but the chatbot wasn’t adapting to the nuances of each query. I had built a tool that could answer questions — but not one that could understand them.

I revisited the chatbot’s training data and realized I had missed something crucial. I wasn’t just training it on answers, but on the context of the questions. I needed to teach the AI not just what the answer was, but why it mattered. That meant refining the training data to include more detailed examples and ensuring the chatbot could interpret the intent behind each question.

This change didn’t just improve the chatbot’s performance — it also changed how I approached building AI tools. I learned that the most effective AI systems aren’t those that just repeat answers, but those that understand the questions behind them.

This experience was a turning point. It taught me that building AI tools is not just about writing code — it’s about listening, understanding, and asking the right questions. It’s a lesson I carry with me in every project I work on, whether it’s building an e-commerce store, a language learning app, or a planning assistant.

The chatbot that said 'no' didn’t just fail — it opened the door to a better way of thinking. And that, in the end, was the most valuable lesson of all.