Working with the latest self hosted docker n8n with the template below
Chat with OpenAI Assistant (by adding a memory)
Work case:
we have it currently setup as a basic lead generation flow for a plumber, qualify with progressive information capture in a conversation flow.
Conversation journey configured from Google assistant
Problem, 2) contact details, 3) detail confirmation and 4) confirmation. 5) direct CTA for immediate contact via twilio phone number if required
Question:
How can you setup a trigger in the workflow to monitor critical moments in the conversation to discard unqualified, escalate to live chat or complete the conversation to push to another workflow to submit to the CRM/lead alert etc.
Looking for a more complete example of best practices
Approached it like this, using an IF statement and having the openai assistant instructed to use a very specific phrase when summarising the enquiry and closing out the conversation.
I feel there should be better way of approaching this
I typically work on a objective, process and outcome with exceptions when building these out.
Story building of who, what, when, why and how
Feels like the chatbot has no material outcome on the business side and is having a chat for the sake of it.
We are instructing the assistant separately but no real monitoring in n8n.
Needs to have some sort of data store of details captured in the natural flow of conversation, and then a timeout or some means of monitoring when a chat is abandoned.
Other chatbots I have run the flow below
Lacking the ability to reliably do 4 - 6 in n8n
Capturing contact details (who are you and your contact details), may come before or after the identification of the problem or purpose of the chat. Details are logged on file
Tease out the purpose of the chat - "“how can we help?” → “I am looking to do xyz”
Logical feedback, refinement and problem qualification (lead generation)
Business objectives, push them towards a business objective? Product Recommendations (ecomm), Call back (speak to a senior member), book an appointment (calendly)
(send data to xyz/action xyz)
Summarise the conversation, confirm satisfaction and close off the chat.
(send data to xyz/action xyz)
If form is abandoned, 2 min check in on chat - “are you still with me?”
Summarise the conversation and data (send data to xyz/action xyz)