Overview

The AI Agent module in Milestone’s Platform allows agencies to create, configure, and deploy AI-powered chat agents on their client websites. These agents can respond to customer queries, use brand-aligned conversational tone, and draw from structured knowledge sources such as your website, uploaded files, and Q&A entries.

This guide explains how to:

• Tool and enable AI Agents for a business
• Create and configure an AI Chat Agent
• Customize the chat interface
• Manage knowledge sources
• Test, preview, and deploy the agent

Tooling and Enabling AI Agents for a Business

Purpose

Before you can create or configure an AI Agent, the onboarding team must enable (“tool”) the module for the business in the MPC.

Steps

Step 1: Go to the Business Section in MPC.

Step 2: Open the Tooling tab and locate the AI Agent module. Enable the AI Chat Agent for the business.

Once enabled at the business level, the AI Agent module becomes visible at the profile level.

Note: The AI Agent module is disabled by default and must be manually enabled by the onboarding team for each required profile. Once enabled, the AI Agents Dashboard becomes accessible for that profile.

Creating a New AI Agent

Step 1: In the left navigation panel, click on AI Agent

Step 2: Select the profile for which you want to create an agent.

Step 3: Click Create Agent.

Step 4: Provide a name for the Agent; this is an internal identifier.

Step 5: Select the Agent Type.

You will now be redirected to the AI Agent Settings section.

Configuring Agent Settings

The Settings page controls how your agent behaves and communicates. Enter the information to start training the agent. 

Provide Agent Information.

• Agent Name: Enter an internal reference name (not visible to end users).
• Description: Provide an instructional prompt that defines the agent’s role and behavior.
  Example: “You are a support agent for my brand (name). Provide concise, friendly responses about vacation rentals."

• Brand Voice: Set the tone and personality for your agent’s conversational style. This is distinct from CMS/Content Studio voices and limited to 500 characters for clarity.
• Delete Agent: Permanently remove the agent and its configuration 
  Note: The brand voice for chat focuses on conversational tone, not editorial writing style.

Customizing the Chat Interface

The Chat Interface tab lets you personalize the look, feel, and user experience of your AI Chat Widget. 
Note that as you make updates to the settings, you will see the live preview of your chatbot (on the right of your screen) update in real-time.

Click on the Chat Interface tab at the top of this section.

General Settings

Chat Icon: Upload the chat icon that will appear on your website. Follow the size and image file requirements.

Agent Profile Picture: Add a display image for your agent. Follow the size and image file requirements.

Agent Name (Display Name): Specify the visible name users will see (different from the internal agent's name). Maximum of 50 characters.

Welcome Message: Set the initial greeting your agent sends, for example, “Hi! How can I help you plan your stay today?”

Action Chips: Create up to six predefined conversation starters, such as “Find Properties,” “Check Availability,” or “Contact Support.” Users can click these to begin a chat.

Tip: Action Chips are fully customizable, you can add, remove, or reorder them as needed. Each chip simply sends the exact message you define, with no hidden logic.

Message Placeholder: Define the text shown in the input box before typing, such as “Type your question here or Ask anything”.

Response Feedback: You can choose to enable/disable the response feedback. Enabling will display a thumbs-up or down button for users. The responses will help train the agent for future queries.

Design Settings

The Design Settings section controls the visual appearance of your chat widget.

• Primary Color: Applies to the chat header, send button, and hover states.
• Secondary Color: Used for borders and subtle interface elements.
• Font & Background: Currently standardized with a gray base theme.

Compliance & Footer: Optionally add Privacy Policy or Terms links to display in the chat widget footer.

Testing and Deploying Your AI Agent

Playground

The Playground allows you to preview and test your agent before making it live. Ensure that you play with the agent thoroughly before moving to the deployment phase. 

• Preview: An interactive sandbox where you can chat with the agent in real time and verify its responses.
• Deploy: Use this mode to go live by embedding the chat widget code into your website.

\

Deployment Instructions

• Click Deploy to view the Embed Code.
• Copy the HTML snippet.

• Add it to the <head> section of your website pages where you want the agent to appear.

Once added, toggle the Agent Status to “Live” within MPC.

• When “On,” the agent is active on your website.
• When “Off,” the chat widget is hidden (useful for temporary disablement).

Note: The embed process follows a standard web script pattern, similar to integrating analytics or tracking tools.

Sharing for Review (Optional)

• Use the Share option to generate a public preview link.
• Allows users without platform access to test and review the agent.

Publish

Currently, all changes made on the platform are immediately live.

Note: In the next phase, Publish will be required to push changes from draft to live state, allowing controlled version management.

Managing Knowledge Sources

The Knowledge tab defines the information base your AI Agent uses to respond.

Default Website Crawl

When the agent is tooled for a profile, MPC automatically crawls the website and creates an internal knowledge store. If not, enter the URL and Start Crawl. This forms the base dataset for the agent.

Adding Knowledge Sources

You can extend your agent’s knowledge by adding different types of content:

• Links: Add specific URLs such as blog pages or FAQs 
• Files: Upload supporting documents in formats like .pdf, .docx, .pptx, or .txt. Each file can be up to 10 MB, and multiple uploads are supported.

Q&A: Manually add question-and-answer pairs to guide or override the agent’s responses for specific queries.

File Upload Rules

• Each file ≤ 10 MB.
• If any file exceeds 10 MB, it won’t upload.
• Upload status shown as In Progress, Completed, or Failed.
• Supports bulk upload, deletion, and downloads.

Managing Entries

• Use multi-select to delete or download multiple files.
• Use Q&A to directly control responses for certain customer questions.

Agent Dashboard Overview

Once your agent is set up, you’ll see:

• A tile view of all agents for the profile.

Each tile links to:

• Settings (Configuration)
• Chat Interface
• Knowledge
• Playground (Preview)
• Deploy (Live settings)

Troubleshooting & Best Practices

• Agent not visible on website: Check that the embed code is correctly placed and the Agent Status toggle is turned ON.
• Upload failed: Make sure the file size is under 10 MB, and the format is supported.
• Incorrect tone or answers: Review the Brand Voice settings, Knowledge links, files or the Q&A entries.
• No responses in Preview: Ensure the website crawl has been completed and that knowledge sources have been added.

Summary

Setting up an AI Agent involves three key steps:

• Enable the Agent module for the business and profile.
• Create and configure your agent’s tone, interface, and compliance settings.
• Add knowledge sources, test via the Playground, and deploy it live.

Once deployed, your AI Chat Agent becomes a dynamic extension of your brand - ready to engage visitors, answer questions, and drive conversions.