Or, the longer your chatbot “lives” the more it can learn. If you started small (the sensible route to launching a bot), then you can monitor and analyze the most common requests to your chatbot and create a training plan aimed at building up the bot’s knowledge and broadening the number of questions it can easily answer.
However, in the case of chatbots, failure actually IS an option. In fact, it’s one of the key ways that your bot will evolve and improve. When a chatbot is unable to answer a question, in addition to offering a suggestion and/or sending the user to a human for help (as GeckoBot does), the back end of your bot system will alert you that a low confidence answer was offered. Check this frequently, accept answers that the bot got right into the answer bank and add new answers for the ones it got wrong, or only partially correct. Doing this daily grows your chatbot’s capability and helps it to learn.
On that note, how do chatbots “learn”? Aside from the training that you provide, chatbots are powered by Artificial Intelligence (AI), so have the ability to learn on their own, too. The term used to describe this is “machine learning”, a sub-category of AI that enables a chatbot system to improve from experience without being explicitly programmed or trained. In essence, the chatbot recognises patterns over time and uses these to make better decisions in future when asked similar questions.
When it comes to learning, between your efforts and the chatbot’s own you will have a smarter bot with every day that passes.