Note: This feature is currently in private beta. Contact your CSM or write to support@freshdesk.com to have it enabled.

The Skill Builder is a no-code interface that lets Admins create Agentic workflows—called Skills—for AI Agents, without writing a single line of code. Each Skill is built using modular blocks that define how the AI Agent should:

• Collect information from users
  
  
• Evaluate conditions
  
  
• Take actions across different systems
  
  
• Respond with the right information
  
  
• Or hand off to a human Agent when needed

Here’s a breakdown of the six key blocks available in the builder:

Trigger Block

What it does:
This is the starting point of any Skill. It defines when the AI Agent should initiate a Skill, based on user input.

How to use:
Click the Trigger block to open a configuration panel where you can enter natural language instructions. This is where you define phrases or expressions customers might use to trigger the Skill.

You can optionally include real-world examples and business-specific terms to make the AI Agent more effective. For instance, in an e-commerce domain, users may use terms like:

• “Check RTO status” (Return to Origin)
  
  
• “COD order issue” (Cash on Delivery)
  
  
• “Where’s my tracking number?”
  
  
• “Exchange window expired?”
  
  

Providing these domain-specific examples helps the AI better recognize customer intent and respond accurately—especially for industries with nuanced workflows or unique vocabulary.

Best Practice: Use varied, natural-sounding examples your customers actually say.

Note: A common misconception is that the Trigger block can act on system events (like conversation initiation or positive/negative feedback) or detect sentiment. It cannot. The Trigger block only responds to customer queries.

Collect Info Block

What it does:
 This block enables the AI Agent to ask for and store information from the customer in temporary Skill-scoped variables.

How to use:
 For each variable, use a simple prompt like:

“Ask the user for their 6-digit order ID”
 “Collect the email linked to the subscription”

Along with the prompt, you can add a description to clarify what the property is about. For example: if the parameter is email address and the description is email address associated with the booking, the AI Agent will procure the email address used for booking.

You can then configure each variable as:

• Mandatory – The AI must collect this before proceeding. It will keep prompting until the customer provides it.
  
  
• Optional – The AI will try to collect it, but will move on even if the customer doesn’t respond.

There are four types of Collect Info:

1. Text - Collects free-form input such as names, addresses, or order IDs.
   
   
2. Number - Collects numeric input such as quantity, PIN codes, or ticket numbers.
   
   
3. List – If you define items in the description, the AI Agent will still treat the customer’s response as text input. The description simply helps guide the response, but won’t display as a selectable list.
   
   
4. Boolean – This will be rendered as plain text (not as a radio button or toggle).

Where it’s used later:
 These collected variables can be used to:

• Set conditions in a Condition Path
  
  
• Fill fields in an API Action
  
  
• Personalize messages in Custom Responses
  
  

Best Practice:

• Keep questions concise and relevant
  
  
• Make only essential fields mandatory to reduce friction
  
  
• Use terms your customers are familiar with (e.g., “tracking ID” vs. “shipment reference”)
  

API Action Block

What it does:
 This block connects your Skill to external systems via APIs, so the AI can retrieve or send data.

Two parts to configuring this block:

1. Choose or Create API Action

• Select an existing API action from the dropdown

Note: API actions you've used across all skills will be available here to choose from.

• Or, click Create new API action to configure a new one from scratch.

To create a new API:

• Enter a name
  
  
• Define input variables
  
  
• Configure the API URL, headers, auth, and body
  
  
• Test the response
  
  
• Define output variables to store and reuse later
  
  

Note: Once created, API actions are reusable across multiple Skills.

2. Map collected inputs to API parameters

Map your collected variables (e.g., order ID, email) to the corresponding API fields.

Under Advanced preferences, you can enable Ask for confirmation. When turned on, the AI will confirm with the customer before performing any sensitive action—like canceling an order or changing a plan.

Best Practice:
Always test your API action during setup. Use confirmation toggles for high-impact operations.

Condition Path Block

What it does:
This block allows the AI Agent to branch into different logic flows based on information collected or retrieved earlier.

You can:

• Set multiple conditions (e.g., if order status is “shipped”)
  
  
• Add an Else path as a fallback
  
• Use outputs from the API or collected variables in your logic
  
  

For example, after showing order details, use a Condition Path to check:

• Has the item shipped already?
  
  
• Is the order eligible for expedited processing?
  
  

Based on the answers, the Skill can continue with expedited handling or inform the user why it can’t proceed.

Best Practice:

• Keep each path focused and purposeful
  
  
• Add fallback paths for incomplete or unexpected responses
  
  

Custom Response Block

What it does:
This block allows you to manually define how the AI should respond to the customer.

You have two options to set up your responses:

• AI-generated response - Use when the reply depends on customer inputs or fetched data (e.g., “Your flight from {origin} to {destination} departs at {time}”). Saves time and ensures dynamic accuracy.
  
  
• Send your own response - Use when the reply must always stay the same (e.g., “Please contact support for booking changes”). Good for policies, disclaimers, or standard instructions.
  
  

Best Practice:
 Use Custom Responses when tone, brand voice, or exact phrasing matters.

Agent Handoff Block

What it does:
This block transfers the conversation to a live human agent when needed. This is useful in a workflow or specific condition requires handing the conversation over to agents.