The chatbot does not understand: an effective chatbot will manage the context and will correctly answer the questions that are present in the knowledge base; also it will be able to make specific actions, at the right point of the conversation flow.
However, it is possible that the bot could not understand a request correctly. Why does this happen? Let’s take a look at some possible reasons.
Knowledge base is not populated: the chatbot does not understand an unexpected request in the knowledge base, but also an alternative way to ask that given question. If this happens, the bot returns a message of non-understanding, which will be reported among the “misses” (un-handled requests), within the “Improve” step of crafter.ai platform.
Alternatively, if the bot is in doubt about the correct answer to give, it will return disambiguation alternatives. This means that it will show buttons containing possible interpretations that the user can choose from. If one of the button will match the domain of the user’s request, the conversation will proceed smoothly.
The bot goes into conflict: the chatbot may not correctly answer a question included in the knowledge base. This happens because if two different questions have the same answer, the bot gets confused about what answer to give. To prevent this, the crafter.ai platform flags variants of a question that may possibly get “in conflict” with others already entered. However, to avoid conflicts you better make sure to type specific and distinct answers for each single Q&A.
Questions that require an additional flow: the bot’s behavior depends on the conversational flow that it is able to handle. A simple Q&A automation bot will answer all the expected questions, but will not be able to make actions that involve the management of specific skills, such as, for example, collecting user data for lead generation, or handing over the conversation to the operator.
If you need your bot to accomplish specific goals, those must be included in a specific pre-designed conversational flow.
For example, Crafter.ai platform integrates functionality “bricks” that allow you to add specific capabilities to your bot.
Why is the chatbot offline?
After launching the training, or after making changes to the chatroom settings and design, the Crafter.ai platform prompts you to update the bot.
Updating the chatbot takes a few minutes, during which the bot will not be accessible to users who will attempt to interact with it in that time frame.
For this reason, it is advisable to make updates during times when there is less traffic on the channels that host your bot.