Who else is facing inconsistency with performance? I love what’s being created here, but it feels too unstable to be called reliable. Often, things suddenly stop working properly. Fetching data from the knowledge base is particularly unreliable — sometimes the assistant accurately follows the KB, but other times it’s completely off. It might even say the opposite of what’s in the KB or claim it doesn’t have access to information that is clearly part of the KB.

Is there a way for the AI to wait about 15 seconds before starting to draft a response? It often happens that contacts send two messages almost simultaneously, and unfortunately, the AI ends up sending two consecutive messages as well. If you're recommending the wait time... it doesn’t work correctly.

Is there a way to revert to a previous prompt after saving it, similar to the history log in GHL?

Is there a way to configure AI to capture multiple user responses before replying? Here’s an example: when the AI asks for contact details, users might send each piece of information in separate lines or messages. I want the AI to wait until it receives all the information before responding. Another example is when a user sends multiple questions in separate messages one after another. The AI should be able to capture all the messages first and then provide a comprehensive response. Has anyone figured out how to achieve this already?

Is anyone else learning this platform and would like to meetup over zoom and discuss together?