One of the automations in n8n with
# 🧶︱general
r
One of the automations in n8n with several examples, which explained above that it used Pinecone as its Vector DB and that it was very good since it recommended it. I want to create a chatbot for the support lines since I have seen that it still uses the typical chat forms where you have to choose from the options, and I feel that it is inefficient since, in the end, it does not really resolve the person's questions, and if it does not resolve them, they pass the case on to a person, and this takes hours or is not even addressed. This forces people to go to customer service at a branch, and if it's a silly question that could be resolved quickly, it's a waste of time just to ask about a product or service they are providing.
p
ok, very cool. @fierce-businessperson-64086 might have some patterns for you for support chatbots. He created something recently that demonstrates the ability to understand what exactly is the problem and gather the necessary details before answering the question
f
Hey @red-apartment-3953 , sorry for the late response. The main reason support bots feel useless is they answer too early without enough context. A better approach is to first confirm what the user is actually trying to do and collect only the missing details with 1–3 targeted questions, then retrieve the relevant internal docs/policies from Pinecone and answer with citations so it’s grounded and auditable. If it still can’t resolve, it should escalate with a clean case bundle (summary + gathered details + what it looked up) so a human can finish fast without restarting the conversation. If you share your industry, the top 5 question types, what knowledge sources you have (FAQs/KB/PDFs/tickets), and whether the bot needs account-specific lookups, I can try and help more. I would also love to know where the automation fails - that will help me understand the issue you're running into.
r
What you've mentioned is exactly what I want to do. For now, I’m not looking to specialize in a specific niche; instead, I want to start with something generic. Although I’m passionate about AI, Vector Databases, and the endless possibilities they offer, my knowledge is purely theoretical and high-level—enough to understand the concepts, but not quite enough for a deep technical conversation. As I mentioned, I want to build a generic chatbot. As a user, it’s frustrating to reach out to support only to encounter a bot that forces you through a series of automated options. Navigating these menus without resolving the issue inevitably leads to human escalation, where response times take hours, or the ticket is simply closed without being addressed.
I want to allow users to state their requests in plain language, having the chatbot respond using data from a Vector Database. Only if it cannot answer the question should it escalate to a human agent. My motivation comes from my own experience: I couldn't find information regarding transaction times or failure protocols, and because the bot couldn't answer, it kept pushing me to a human representative. It felt like a waste of time for everyone, especially since I eventually found the answer myself online. I might have rambled a bit, but I wanted to provide context so your guidance can be more focused on my specific goals.