Overview
Once you have configured and tested your AI agent, the next step is to deploy it on WhatsApp so it can start handling real customer conversations. Charles offers two flexible deployment options to suit different use cases and rollout strategies: the Global AI Agent and the Journey Flow Agent. This article explains both options to help you choose the right approach for your business.
Deployment Options
Option 1: Global AI Agent
What it is: The Global AI Agent is available across all incoming WhatsApp conversations. When activated, it replaces the catch-all flow, meaning the agent becomes the default handler for every message your customers send.
Best for:
Businesses ready for a full-scale AI rollout
Use cases where the AI agent should be the first point of contact for all customer inquiries
How to activate:
Navigate to Charles AI Squad → Agents and select your agent.
Choose the Global AI Agent deployment option.
Activate the agent — it will now handle all incoming WhatsApp messages by default.
⚠️ Important: Activating the Global AI Agent will replace your existing catch-all flow. Make sure your agent is thoroughly tested before enabling this option.
Option 2: Journey Flow Agent
What it is: The Journey Flow Agent is deployed within a specific Journey Flow, making it available only at a defined point in the conversation — for example, after a customer selects a particular menu option or reaches a specific step in a flow.
Best for:
Businesses that want to test the AI agent with a subset of conversations before a full rollout
Use cases where AI assistance is only needed at a specific stage of the customer journey (e.g., product discovery, post-purchase support)
💡 Tip: Starting with a Journey Flow deployment is a great way to validate your agent's performance in a controlled environment before going global.
How to Set Up Your AI Agent in a Journey Flow
Setting up your AI agent in a Journey Flow is straightforward. Follow these steps to get your agent live:
Upload your Knowledge Base or Product Catalogue to your Universe.
Configure your AI agent(s) in Charles AI Squad.
Create a new conversation in the Journey Flow builder.
From the drop-down menu of available nodes, select the AI agent node — "Charles AI".
In the left-hand panel, add your configured AI agent(s) to the node.
Configure your product display format (if applicable):
If your AI agent is Recco, our product recommender, products will be sent in up to five free message templates by default.
Alternatively, you can choose to send products in a Carousel Template for a richer, more visual experience.
Configure your Fallback Message — this message triggers only if the user's query cannot be answered by any of the AI agents in the node.
Customise the text message, CTA button, and action trigger within the fallback.
Set a Timeout Limit — this defines the time after the user's last engagement before the follow-up step is triggered.
⚠️ Important: Ensure that the message node directly preceding the Charles AI node contains a custom property that stores the user's input. This is essential so that the agent correctly recognises the user's query. If you have not yet created a custom property, navigate to Custom Property Settings and create one before proceeding.
💡 Tip: It is recommended to introduce your agent to the customer before the Charles AI node — for example, with a short message explaining that they are now chatting with an AI assistant.
💡 Tip: Starting with a Journey Flow deployment is a great way to validate your agent's performance in a controlled environment before going global.
Example Journey Flow Setup
Below is an example of how a Journey Flow with an AI agent node might look in practice.
Choosing the Right Deployment Option
| Global AI Agent | Journey Flow Agent |
Scope | All WhatsApp conversations | Specific point in a Journey Flow |
Replaces catch-all flow | ✅ Yes | ❌ No |
Best for | Full-scale rollout | Targeted rollout / testing |
Flexibility | Lower | Higher |
Recommended for first-time deployment | ❌ | ✅ |
Your AI agent is now live on WhatsApp. Whether you opted for a global rollout or a targeted Journey Flow deployment, you are equipped to monitor, learn, and continuously improve your agent's performance.
Best Practices
🗂️ Save the User's Query
Add a custom property to capture and store your users' queries. This helps you:
Understand what questions and products your users are most interested in
Gain direct insights into how to optimise your FAQs or product catalogue
📣 Add a Follow-Up Step
Use the Timeout Limit to guide users after they have stopped interacting with your AI agent. Whether you direct them to your happy path or collect NPS data, a well-designed follow-up keeps the conversation going.
💡 Tip: Incorporating an NPS score in your follow-up is a great way to collect feedback immediately after a user has engaged with the AI agent. Use this data to continuously optimise your agent's performance.
🏷️ Add Tags
Add tags to your Journey Flow to help you segment your audience and track the performance of specific parts of the conversational experience.
🗣️ Add a "Contact Support" Option in Your Fallback
Avoid user frustration by including a "Get in Touch" CTA button within your fallback message. This button should direct users to your customer support tool or provide clear information on how to contact your team.