This cookbook gives you the exact steps to create, train, and deploy your AI support agent using YourGPT.
You’ll start by creating your workspace, then add your business data, customise your agent’s behaviour and appearance, and optionally build advanced workflows. Each step is designed to help you move fast and get a production-ready AI agent without needing code.
Follow the steps as-is or change them based on your support needs.
🧩 Step 1: Create Your Account
Start by setting up your YourGPT workspace:
Go to yourgpt.ai and sign up.
Start with 7-day free trial
Access your dashboard and launch the AI Studio or No-Code Builder.
✅ You’re now ready to build your first AI support agent.
📥 Step 2: Train Your Agent on Your Business Data
Upload the knowledge your customers need:
Documents: PDFs, DOCX, CSVs, FAQs, SOPs, onboarding guides.
Cloud Sources: Notion, Google Drive, Confluence, Dropbox.
Media & Links: YouTube videos, support tutorials, past chat logs.
YourGPT will ingest and index the data so your AI agent can answer questions accurately and contextually.
💡 Pro tip: Organize your data by support categories (orders, returns, tech help, etc.) using training tags for even better results.
🎨 Step 3: Customize Your AI Agent
Now that your AI has training data, let’s tailor how it acts, speaks, decides and looks.
✏️ Agent Personality
You can fully customise your AI agent personality
Define the tone of voice: Friendly? Professional? Playful?
Add a persona description like:
“You are Ava, a smart, helpful assistant for CR Tech. who always explains things politely clearly and concisely.”Using template functions you do make bot more aware of like time zone, user details & more.
🔍 Define Boundaries
Add restrictions so the AI avoids unsupported topics.
Configure fallback responses like:
Do not answer I’m not sure about that — let me connect you to a human!You can use our Generate with AI to create persona for your needs.
⚙️ Other Settings
Pick the right AI model for your need — check our recommendation if you're unsure.
Adjust temperature for more creative or precise answers.
Limit previous message memory for shorter or more focused responses.
Connect to specific Knowledge Base nodes to guide its answers.
🎨 Customise Appearance
Your AI widget should look like part of your brand, not an add-on. Here's how you can customise its appearance:
Theme & Colours:
Match the widget’s colour palette with your brand's primary and secondary colours. This includes the background, button colours, and text.
Logo Placement:
Upload your brand logo and decide where it appears – header, welcome screen, or chat bubble.
Font & Typography:
Choose fonts that align with your website’s style. Use size, weight, and spacing settings to maintain consistency.
Welcome Screen Text:
Write a greeting that feels on-brand. Example: "Hi! I’m Ava from CR Tech. How can I help you today?"
Widget Layout:
You can pick between Tab & Chat Layout.
Select the position on the screen – bottom right, left, or full-page overlay.
Decide if the widget should open by default or stay minimised until clicked.
Interactive Elements:
Add branded buttons and icons for actions like starting a new conversation, escalating to a human, or accessing FAQs.
Customising these elements ensures that every interaction with your AI agent feels native to your website and familiar to your users.
If you need you can also add a AI search widget.
🎯 Goal: Ensure your AI reflects your brand's personality, operates within defined boundaries, and integrates seamlessly as an extension of your team.
🧠 Step 4: Design Sequential Agents (for advanced Needs)
For structured, multi-step workflows, you can use AI Studio to build sequential agents. This gives you full control over AI agent. It's optional—but useful when you need custom behaviour.
Components of Sequential Agents
📌 Intents
Intents capture the user's purpose or goal. They link user queries to specific actions or responses.
Example: User asks, "Book a flight." Intent = "Flight booking."
🚩 Events (Trigger Points)
Events initiate a conversation flow when specific actions occur:
New Session: When a new user session begins.
Request Human: When the user asks to speak to a human.
Unable to Answer: When the AI can't provide an answer.
Contact Created: When a new user contact is added.
AI Response Follow-Up: After the AI responds from the knowledge base.
Session Resolved: When a session is marked as resolved.
🗂 Entities
Entities extract critical details from user input, like dates, locations, or names.
Example: "Book a flight to Paris on March 5th." Entities = "Paris" (destination), "March 5th" (date).
📋 Variables
Variables store user information to personalize interactions and maintain context across messages.
⚖️ Conditions
Conditions allow your chatbot to make decisions based on user input or stored data:
AI Condition: Uses AI understanding to route conversations.
Manual Condition: Uses specific rules to guide conversations.
Suitable for precise logic (e.g., exact matches, numerical comparisons).
💬 Messages
Messages are the content your AI bot can sends to users:
Text: Simple, informative responses.
Image: Visual content to engage users.
Video: Dynamic and interactive content.
Audio: Send audio clips.
File: Share documents and structured information.
Card: Combines visuals, text, and actions.
Carousel: Interactive scrollable sets of cards or images.
🤖 AI Response
AI Response Generate dynamic responses based on user input and conversation context:
AI Response: Uses built-in general AI knowledge to generate responses.
Customize response instructions and formats (Text, JSON).
Knowledgebase Response: Uses your business-specific knowledge base to answer user queries accurately.
Customize using tags, response formats (Text only), and prompt settings.
🧐AI Retrieval:
It retrieves directly from your knowledge base. It only retrieves, it do not generates response You can use this data to generate structured JSON and display the data as a carousels.
⏳ Listeners
Listener allows you to Capture user input, button, forms, feedback.
✅ Action
Actions are Operational tasks that your bot can perform within a conversation:
Escalate to human agents
Transfer between bots
Update conversation segments
Mark sessions resolved
Create new contacts
Send automated emails
🌐 Webhooks
Integrate external systems or data directly into your chatbot flows.
⚙️ API Calling & Code Execution
Call API's or execute custom Python, JavaScript for dynamic data processing.
🔄 Components (Reusable Blocks)
Components are Reusable sets of nodes:
Select nodes → Right-click → "Create Ccomponent."
Simplifies and standardizes chatbot development.
We've created a complete template to help you quickly get started with AI Studio and build custom AI workflows for support. Download it here.
📚 Step 5: Create Your AI Helpdesk
A self-service helpdesk empowers your customers to quickly find answers on their own—either through direct browsing or by asking your AI agent. It also serves as a valuable self service knowledge source that helps your customers, internal team and improves your AI’s understanding of your business.
🔧 How to Set Up an AI Helpdesk:
Add a Workspace Name
Create a dedicated workspace specifically for customer support helpdesk content.Choose a Helpdesk Template
Pick a layout that fits your needs—simple, FAQ-style, or document-rich formats.Create Articles
Add helpdesk content manually or generate it using AI. You can include:Plain text
links
Images
Audio clips
Publish and Share Your Helpdesk
Share it via a direct link
Or embed it into your website or AI chat widget
💬 Step 6: Use the Conversation Dashboard
YourGPT’s Conversational Dashboard is the command center for your support operations. It’s where your AI and human agents work together to deliver fast, consistent, and accurate support.
AI Handles First Response: AI replies instantly using your training data and workflows.
Human-in-the-Loop Escalation: When the AI can’t answer or detects frustration, it escalates the chat to a human agent.
Smart Session Data: From the Right hand pannel, view visitor info (Contact info, location, browser, device) and session insights (IP, last message, tags).
Quick Actions Panel: Resolve tickets, escalate to agents, request feedback, or export the chat.
Unified Inbox: Manage conversations from Instagram, WhatsApp, Messenger, Crisp, and more—all in one place.
Filters & Views: See chats by channel, tag, agent, or status (assigned, resolved, awaiting or create your custom views).
Real-time Translation: Instantly translate incoming and outgoing messages for multilingual support.
Auto-Summary & Feedback: Quickly review what happened in a chat and get customer satisfaction insights.
If you need any further more help, join our Discord community at https://discord.com/invite/z8PBs5ckcd to get direct support from our technical team.