Total newbie here, looking to get some pointers to get me going in the right direction
I´ve followed this tutorial on how to create a RAG chatbot and i´ve got it working. I now want it to determine whether the person chatting with the bot as intent of buying something or just want to chit-chat with the bot Is it even possible?
If the visitor shows any buying intent, I want it to go in to a certain lead-workflow with certain questions there.
Not looking for anyone to solve this for me, I just need someone to push me in the right direction so I learn this.
Thanks!
for your idea to work you have to provide proper prompts to the ai agent that would discerne the intent of the person it is chatting with , try watching this to understand more how to build your prompts
However, i´m pretty sure I´ve got the prompts alright. It´s more an issue on how to set the flow up, how do divide them etc.
Should the “main chat” be one flow that sends the visitor into another flow when purchase intents are noticed?
My idea is basically like this:
Visitor talks to the chatbot, chatbot determines whether the visitor has buying intents or not. Yes, go on asking specific questions about this and summarize the answers in a mail. No, keep the conversation going and keep checking for purchase intent.
Do i make any sense?
I´m pretty sure it can be achieved, i´m just not sure how to set all this up.
aok, I’m not entirely sure about what you mean.. I was just in the impression that based on your first message you were asking if it was possible for the AI to discern the intent that’s why i gave the direction of how to set the prompts as that’s how you can instruct the agent about its role and temperament.
I’m guessing that since you already got the prompts right then what you are missing is tools for the AI to use. Thats where you will establish the email then the next part of the flow is your email service to send it.
with regards to specific questions or keeping the conversation going, again that would reside on your prompts . It would be good to set up examples and rules in your system prompts so that the agent can lead the conversation based on how you want to lead the conversation.
Correct me if I misunderstood but do you mean that if it determines intent, it would trigger a different node? and if not it keeps checking?
Also I suggest you figure out which is the best model for your use case. Hope this helps!
Is it even possible?