Customer chatbot with sequential questions

I am building a customer chatbot that should have a flow or should ask questions (fixed questions) in a sequence so that when all questions are asked we can raise a ticket if issue is still unresolved.
The bot’s first answer is for a hardcoded question i.e., where is my item?
then when the bot tells that the item has been delivered on so and so date.
The user can either say ‘i received it’ or ‘ok received it’.
Depending on that bot responds with ‘can you confirm the address?’ if user did not received it, and ‘thank you!’ if user gets it.
Then there is sequence of questions like a tree, that need to be asked depending on user’s answer.
In the prompt, I am providing the ‘chat history’, all the instructions on what to respond when user replies in a certain way to a previous question, ‘last ai response’ separately to make it more clear.
Instructions are like:
Use ‘ai response’ to build context. Always follow the following instructions:
→ To your question ‘is this your house?’, if user then says yes its my house or something like - ‘yes’, then respond saying something like - ‘can you check at home?’. But for the question ‘is this your house?’, if user then says not my house or ‘no’ then reply saying something like - ‘check with neighbours’…(sequence of questions go on depending on human answers)

The problem I am facing is that sometimes if user says ‘no’ then it asks the wrong question because ‘no’ can be an answer to many questions in instructions. Another problem is that when user gives a random reply, the chatbot doesn’t include the flow in the answer or doesn’t try to go only with the flow.

It actually sounds like you’d be faster / better off using an old-school ‘scripted’ fixed-question-fixed-answers chatbot that runs on a decision tree, not a large language model. Feels like a bit overkill for the information you’re requiring, which - as I read it - could also be attained using a typeform of google form. Or am I missing the point here?