Maintain a chatbot in production
When your chatbot is live for your users, use Oswald built-in tools to test and analyse your chatbot. You can use these test results and analytics to extend and improve your chatbot.
When your chatbot is live for your users, use Oswald built-in tools to test and analyse your chatbot. You can use these test results and analytics to extend and improve your chatbot.
When your chatbot does not behave the way you want it to, you can take a look at the Oswald bot brain. You can literally look inside the brain of the chatbot, see how it interprets the user input and why it chose a specific answer. The bot brain gives you a clear overview of what is happening inside your chatbot.
Tests in Oswald are samples of conversations with your chatbot. After you made some updates or changes, you can simulate all these conversations with your chatbot with the click of a button. The results are pretty clear: when all tests are green, your chatbot still works as you expect it to. When some tests are red, something is wrong and needs to be fixed.
Oswald gives you advanced analytics for your chatbot. Choose a timeframe and you get the number of messages, the number of sessions, the average response time, the number of human takeovers and the number of unaswered questions. Furthermore, you get an overview of the length of the conversations and you see which intents and scenarios are triggered the most.
Questions from users that did not get a response from the chatbot are logged in Oswald. You can input the question in the chatbot and see why the question was unanswered. If the intent was not recognised correctly, you can easily select the right intent and validate with the click of a button. Your chatbot gets smarter over time.
