Introduction

Watching the documentary Startup.com, I was struck by how one of the most challenging parts of starting a business in the 90s was coming up with a catchy domain name. It was the peak of the dotcom boom, a time when many companies rushed online, producing little to no real value. Everyone wanted to have a piece of the internet, and having a cool domain name seemed crucial. But in all the excitement, the real question often went unasked: what do you actually do as a business? In hindsight, this should have been the first question on every investor's mind. But in the race to be part of the internet gold rush, many seized any opportunity, however vague, without thinking it through.

This reminds me a lot of the chatbot craze. Around 2018, adding a chatbot to your website became the new "must-have," even if it didn't provide real value. Everyone wanted to ride the wave, fearing they’d miss out on the next big thing. But what was the opportunity, really? Much like the rush for catchy domain names without clear business models, many rushed to deploy chatbots without understanding their purpose or potential.

I’ve seen many chatbots come and go, driven more by hype than by utility. At one point, I joined a small startup as the first engineer. We built what we hoped would be an innovative chatbot, or more accurately, a digital customer service agent. By the time I left two years later, we had grown from 3 to 40 employees and resolved tickets from millions of customers. We learned a lot about what makes a chatbot work—and what doesn't.

Now, with the benefit of hindsight, we can reflect on what makes a good chatbot. What are some of the assumptions that we left behind? If I were to build one today, what would it look like? Let's dive deeper into these questions.