---
title: "Discussion on \"Building a support chatbot a US business can trust: retrieval, guardrails and human handoff\" | Hashnode"
description: "Discussion on \"Building a support chatbot a US business can trust: retrieval, guardrails and human handoff\". A US business owner asks for \"a chatbot on the website that answers customers\". The quickest build is to connect a large language model to the site and let it answer. It demos well. Then, within a wee"
url: "https://hashnode.com/posts/building-a-support-chatbot-a-us-business-can-trust-retrieval-guardrails-and-human-handoff/6af06676825730011571c9f6"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Building a support chatbot a US business can trust: retrieval, guardrails and human handoff](https://dzimarkdev.hashnode.dev/building-a-support-chatbot-a-us-business-can-trust-retrieval-guardrails-and-human-handoff?utm_source=hashnode&utm_medium=feed)

A US business owner asks for "a chatbot on the website that answers customers". The quickest build is to connect a large language model to the site and let it answer. It demos well. Then, within a wee

3d ago · 4 min read · dzimarkdev.hashnode.dev  
Read the article

-   [#chatbot](https://hashnode.com/tag/chatbot)
-   [#llm](https://hashnode.com/tag/llm)
-   [#rag](https://hashnode.com/tag/rag)

## Written by

[Dzimark](https://hashnode.com/@dzimarkdev)

@dzimarkdev · Mohali, India

  
[View profile](https://hashnode.com/@dzimarkdev)

Website, app and AI development for growing brands in the USA, UK, Canada and Australia.

Dzimark is a website, mobile app and AI development company for businesses in the USA, UK, Canada and Australia: Shopify and WooCommerce stores, React and Next.js websites, iOS/Android/Flutter apps, AI chatbots, AI agents and automation. Offices in Dallas (USA) and Oshawa (Canada). https://dzimark.com/

## Responses 1

Markdown supported · ⌘↵ to post

[Ahmet Özel](https://hashnode.com/@ahmetozel)

[3d ago](https://hashnode.com/posts/building-a-support-chatbot-a-us-business-can-trust-retrieval-guardrails-and-human-handoff/6af06676825730011571c9f6/comment/6af1555f922148e5e74980d4)

The resolution-rate metric needs an explicit endpoint. A customer leaving the widget, a bot handing off successfully, and a human eventually closing the ticket are three different outcomes. Counting all three as resolved would make it hard to tell whether the chatbot actually answered the question.

I would keep abandoned sessions, bot-confirmed resolutions, and completed human handoffs separate, then connect the last category to the help-desk outcome. That makes the weekly content review more informative: unanswered questions can indicate missing content, while an accepted handoff with no eventual resolution points to a different operational problem.