ReplyCX Documentation
Login
  • ⛩️Welcome to ReplyCX Knowledge Base! 📚
    • ReplyCX Basics
  • ▶️GETTING STARTED
    • Building a Chatbot
    • Testing a Chatbot
    • Channel Configurations
    • Deploying a Chatbot
    • Utility Tools
  • 🧩ACTION BLOCKS
    • Overview
    • Trigger
    • Send Message
    • Collect Input
    • Buttons
    • Carousel
    • Calendar
    • Send an Email
    • Condition
    • Image Carousel
    • Human Handover
    • Slider
    • Collect File
    • Delay
    • Form
    • Flow
    • Code-block
    • Options
    • Jump
    • List
    • Reply Button
    • AI Model
    • SET AI
    • Answer AI
    • AI AGENT
    • Webhook
    • Javascript
    • Send Status
    • Http Request
    • Dynamic data
    • Whatsapp flow
  • 🦾AI Studio
    • Building a GPT Chatbot
    • Knowledge Base
    • Training on historical live chat to generate response
    • Retrain frequency for URL data source
    • Custom Answers
    • Function Call
    • Prompts
    • Tokens
    • Setting up retrain
    • Advanced Crawling Criteria
  • ⛓️Integrations
    • Types of Integrations
    • Service Call
    • Google Sheets
      • Support for “Update Record” in Google Sheets
    • Codeblock
    • Google Calendar
    • Calendly
    • Zoho CRM
    • Hubspot
    • Dialog Flow
    • Events
    • Google Analytics
    • Freshdesk
    • Salesforce
    • Zapier
    • Airtable
    • Public API's
  • 🟢WhatsApp Business API
    • Prerequisites
    • WhatsApp Business API - Meta
    • Using a test WhatsApp Business API account
    • Product catalog on WhatsApp
    • Sync WhatsApp Template
    • Support for Carousel template message
  • Instagram
    • Using the Instagram Channel
  • 💬Live Chat
    • Overview
    • Saved Replies
    • Manage Saved Replies
    • Message status on live chat
    • Generating Response Using AI.
    • Rewriting existing response with AI
    • Labels
    • Managing Labels
    • Qualification details covered during a conversation
    • Settings
    • Filter conversations
    • Conversation History
    • Close a conversation
    • Related / Past Conversations
    • Elements on conversation card
    • Kind of Conversation Status
    • Copy Chat Transcript of a Conversation
    • Customize Live Chat Screen
    • Restart Conversation
    • Blocking Contacts
    • Agent status on live chat
  • ⚙️Troubleshooting
    • Variable Manager
    • Fallback Variables
    • Human Handover Configuration
    • Clone a bot
    • Preffered Image Dimensions
    • Working of Link Tracking
    • Setting up variables using trigger block
    • Availability of agents in Human Handover
    • Creating loop in the conversational flow
    • Requesting Edit Access
    • Cookies
  • 📢Outbound Bots
    • Outbound Action Blocks
      • Delay
      • Send WhatsApp
      • Send SMS
      • Send Email
    • Building a One-Off Campaign
    • Building a Ongoing Campaign
    • Creating WhatsApp Templates
  • 📱Chat Widget Customization
    • Embedding Chat Widgets
    • Customize Chat Widget UI using CSS
    • Display Chat Widget in iFrame
    • Change Appearance
    • Chat widget 3.0
  • 🏦Account Management
    • Manage Teams
    • Manage Teammates
    • Manage Roles
      • Channel configuration Permission
    • Opt Out Management
  • 📊Reporting
    • Custom Reports
    • Contacts Feature Recap
    • Scheduling Contact Report
    • Weekly Reports and Interactions
    • Export a contact list
    • Tracking link clicks on chatbot messages
    • Agent Report
    • AI Agent Report
    • Outbound Report
  • 🧑‍💻 Support
    • Forget Password
Powered by GitBook
On this page
  • Step 1: Creating a Knowledge base
  • Step 2: Build a bot
  • Step 3: Adding the Answer AI
  • Step 4: Adding fallback measures

Was this helpful?

  1. AI Studio

Building a GPT Chatbot

PreviousWhatsapp flowNextKnowledge Base

Last updated 10 months ago

Was this helpful?

GPT chatbots are smart bots that use data to answer questions.

GPT chatbots are very popular right now because of ChatGPT.

ChatGPT has made it easier to ask questions and get answers that are quick, accurate, and to the point.

Here are the steps to create your own GPT chatbot:

As an example, we'll build a GPT chatbot for a website to answer users' questions. If the bot doesn't know the answer, it will provide customer support contact details.

Step 1: Creating a Knowledge base

The key to a GPT chatbot is developing a knowledge base filled with the information you want it to use for answering questions.

Since we aim to answer questions from the website, let's create the knowledge base.

  1. Go to the AI Studio -> Add Knowledge base -> Add data source.

  2. Select the URL as the data source.

  3. Enter the domain of your website.

  4. Hit 'Train' and wait until all URL (s) are being fetched.

Step 2: Build a bot

Next, we need to create a bot that will communicate with visitors and use the knowledge base to find answers to their questions.

Go to the Bot Builder -> Build a bot -> Inbound -> Web.

Step 3: Adding the Answer AI

Following the trigger action block, we will insert the 'Answer AI' action block.

  1. Include a welcome message such as "Hey there! How can I assist you today?"

  2. Select the "Knowledge base" we developed.

Additionally, we will set up the 'Answer AI' block according to our requirements. For this example, I will use the following configuration:

Instructions:

Instructions guide the AI on how you want the answer to be generated. Write a clear and concise instruction.

“I want you to act as a support agent. Your name is "AI Assistant". You will provide me with answers from the given info. If the answer is not included, say exactly "Hmm, I am not sure." and stop after that. Refuse to answer any question not about the info. Never break character.”

All other advanced settings are kept as it is. You can choose to tweak it if needed.

Step 4: Adding fallback measures

The AI won't always be able to answer your question. This could be because it lacks the necessary data or the question is too vague to generate a response.

Therefore, it's crucial to always have fallback measures in place to assist users.

In this example, after the AI provides an answer, we will ask the user if the response was helpful.

Click on the success path and select "Add another block."

Here, we'll add a button block with the message "Did this answer your question?" and the following options:

  • Yes - Ends the flow with a thank you message “Glad we could help!”

  • No - Apologize and share support details like “Sorry we couldn’t answer your question. Feel free to reach out to us or call us for more assistance.”

This ensures that users always have an alternative way to get their questions answered if the GPT fails. You can also extend the failure flow to gather details about their issues and either create a ticket in your support system or notify your team via email.

🦾