AI Agent

Create an AI Agent

Copy page

Building an AI Agent in Chatmo is a guided process that walks you through each configuration step. By the end, you'll have a working agent ready to handle customer conversations — though you'll probably want to add knowledge and test it before turning it loose on real customers.

Before You Begin

There's no special setup required before creating an agent, but having a few things ready will make the process smoother:

• A clear purpose for the agent — Know what this agent is supposed to do. Support? Sales? General FAQ? This shapes every decision that follows.
• Your brand voice guidelines — How does your company communicate? Formal? Casual? Friendly but professional? Your AI Agent should match.
• Sample questions — Think about the top 10-20 questions customers typically ask. These will help you evaluate whether your agent is set up correctly.

Step-by-Step: Creating Your Agent

1. Navigate to the AI Agent Section

From the Chatmo sidebar, click on AI Agent. You'll see a list of any existing agents (if you've created them before) or an empty state with a prompt to create your first one.

2. Click "Create New Agent"

Hit the Create New Agent button. This opens a modal where you'll start configuring your new agent's identity and behavior.

3. Configure Agent Name and Description

On the General tab, you will find the Agent Identity card with name and description fields. Fill in the following:

Agent Name — This is how the agent will be identified within Chatmo. Your team will see this name in conversation assignments, reports, and settings. Choose something descriptive:

• "Support Bot - EN" (for an English support agent)
• "Sales Assistant" (for a sales-focused agent)
• "Maya - Customer Care" (some teams prefer giving agents human names)

Description — A short internal note explaining what this agent does. This isn't shown to customers — it's just for your team's reference. Something like: "Handles general product inquiries and returns for the US storefront."

4. Set Up AI Behavior and Personality

This is arguably the most important step. In the AI Behavior section, you will find the behavior prompt textarea where you write all the instructions that define how your AI Agent behaves — including the language it uses, its tone of voice, its personality, communication style, boundaries, and what to do in specific situations. Think of it as writing a job description for your AI employee.

A strong behavior prompt typically covers:

• Role definition — Who is this agent? What's their purpose?
• Tone and style — How should they communicate?
• Boundaries — What topics should they avoid or escalate?
• Key rules — Any non-negotiable guidelines.

Here are examples of effective behavior prompts:

Example 1: E-commerce Support Agent

Code
You are a customer support agent for FreshMart, an online grocery delivery service.
Your name is Lia. You are friendly, helpful, and patient.

Key responsibilities:
- Answer questions about our products, delivery areas, and pricing
- Help customers track their orders
- Explain our return and refund policy
- Assist with account-related questions

Guidelines:
- Always greet the customer warmly
- If a customer is upset, acknowledge their frustration before offering solutions
- Never make promises about delivery times you can't verify — offer to check instead
- For payment disputes or billing errors, transfer to a human agent
- Keep responses concise — no one wants to read a wall of text in a chat
- Respond in the same language the customer uses

Example 2: SaaS Product Assistant

Code
You are the support assistant for CloudSync, a file synchronization platform.
Be professional but approachable. Avoid corporate jargon — explain things clearly.

You can help with:
- Feature explanations and how-to guidance
- Account setup and configuration questions
- Billing and subscription inquiries
- Troubleshooting common sync errors

You should NOT:
- Provide legal or compliance advice
- Access or modify customer accounts directly
- Share internal company information or roadmap details
- Make guarantees about future features

When you don't know the answer, say so honestly and offer to connect the
customer with a specialist rather than guessing.

5. Configure Welcome Messages

In the Messages card, you can configure the welcome message text and optionally upload a welcome image. Welcome messages are sent automatically when a customer initiates a new conversation. You can set up two types:

Text Welcome Message — The opening line your AI sends when a new chat begins. For example:

"Hi there! I'm Lia from FreshMart. I can help you with orders, deliveries, product questions, and more. What can I do for you today?"

Keep it short and warm. Let the customer know what the agent can help with so they have a starting point.

Image Welcome Message — Optionally attach an image (like a promotional banner, product catalog, or branded greeting graphic) that accompanies the text message. This works particularly well for sales and marketing-focused agents.

6. Set Response Language and Tone (in the Behavior Prompt)

Language and tone are configured directly inside the behavior prompt from step 4 — there is no separate language or tone setting. The behavior prompt textarea is where you write instructions about what language to use, the personality of the agent, and the communication style.

For example, you might include these instructions in your behavior prompt:

Code
Communication guidelines:
- Always respond in English unless the customer writes in another language
- Use a friendly, professional tone
- Address customers by their first name when available
- Keep responses concise and easy to understand

Here are some aspects of language and tone you can define in the behavior prompt:

• Default language — Specify the language the agent should use. Example: "Always respond in English unless the customer uses a different language."
• Communication style — Define whether the agent should be formal, semi-formal, or casual. Example: "Use a friendly but professional tone."
• Greetings — Specify how the agent addresses customers. Example: "Greet customers by their first name."
• Multilingual support — Instruct the agent to automatically detect and respond in the customer's language.

7. Configure Agent Transfer Conditions

In the Agent Transfer Conditions section, you define when and how the AI Agent should hand off conversations to human agents:

• Transfer conditions — Write the conditions that should trigger a handoff to a human agent. For example: "Transfer to a human agent if the customer explicitly asks to speak to a person, or if the same complaint is repeated more than twice."
• Stop AI after handoff — Enable this toggle to make the AI Agent stop responding once a conversation has been transferred to a human agent.
• Silent agent handoff — Enable this toggle so the handoff happens silently without an explicit notification to the customer.

8. Enable or Disable Features

Depending on your use case, you may want to toggle certain capabilities:

Feature	Description	Default
Auto-reply	Agent automatically responds to incoming messages	Enabled
Human handoff	Agent can transfer conversations to human agents	Enabled
Knowledge base	Agent searches your knowledge base to answer questions	Enabled
Tool access	Agent can use configured API tools	Disabled
Typing indicator	Shows a typing animation before responses (feels more natural)	Enabled

You can always change these settings later, so don't worry about getting everything perfect on the first pass.

9. Save and Activate

Once you're satisfied with the configuration:

1. Click Save to store your agent configuration.
2. Toggle the Active switch to enable the agent.
3. Assign the agent to one or more channels or devices where it should handle conversations.

Your agent is now live — but before directing real customer traffic to it, we strongly recommend running through the testing process first.

Tips for a Great Agent Setup

• Start narrow, then expand. Configure your agent for one specific task first (e.g., answering shipping questions). Once that works well, gradually add more responsibilities.
• Write your behavior prompt as if briefing a new employee. Include the context they'd need on their first day — what the company does, what matters most, what to avoid.
• Review and iterate. Your first behavior prompt won't be perfect. After testing (and after some real conversations), you'll find areas to refine. This is normal and expected.
• Don't try to replace humans entirely. The best AI Agents complement your team by handling the repetitive stuff, freeing humans for complex, high-value conversations.

What's Next?

• Knowledge Base — Give your agent the information it needs to answer customer questions accurately.
• Tools Configuration — Extend your agent's abilities by connecting external APIs.
• Testing Your AI Agent — Simulate conversations and validate your agent's responses before going live.