Despite adding information in a structured manner through documents or rich text, the bot sometimes struggles to answer questions accurately. What steps can I take to resolve this issue?

The first thing I would do is, check your information to be sure it is written in a way that minimizes ambiguity to the greatest degree possible. The second thing I would do, is check the instructions for the KB agent to be sure they are as explicit as possible with regard to what sort of output is expected. The third thing I would consider, is breaking up the corpus of information into multiple, narrowly-scoped KBs (scoped by information type or subject) and use transitions to govern which of these narrow KBs is queried. Also, you might want to experiment with different chunk volumes and temperature settings.

How can I change the temperature settings?

Temperature settings are under "Advanced options" in AI Task cards. In short, termperature governs how much variation to expect from an AI output for multiple instances of a given input