Assuming that Lotto Amigo is a game that you want to purchase a ticket for in this story, it seems like your intent classification (Rasa NLU) training didn’t go well. Your bot seems to always identify the message as a “purchase” intent with pretty high probability as well. Can you share your intent training data / configs as well?
What might be the case is that you have many training examples for the “purchase” intent, but haven’t trained the “inform” intent very well. I’m not sure how you’re training the intent to inform about which game you want tickets for, but you probably want to use a generic “inform” intent such as in the weather bot example here: Slot Filling
Unless you have a definite list of games, then you can use lookup tables to increase your “inform” accuracy.