E-commerce Returns Automation

Transform your returns process from a customer service burden into a seamless self-service experience. In this guide, you’ll build an AI-powered returns system that collects return information, validates eligibility, generates return labels, and updates your order system — all through natural conversation. What you’ll build: A complete returns handling system that:

Engages customers via chat, email, or SMS
Looks up their order automatically
Collects return reason and item condition
Validates against your return policy
Generates return labels (via API)
Updates your order management system
Notifies your fulfillment team

Time: 45-60 minutes

Prerequisites

Completed the First Conversational Agent guide
Shopify or order management system connected
Basic understanding of integrations and HTTP requests

The Business Problem

Returns are expensive — not just the refund, but the time your team spends:

Answering “what’s your return policy?”
Collecting order numbers and item details
Verifying return eligibility
Generating shipping labels
Updating inventory systems
Following up on status

With CogniAgent, customers handle their own returns through a friendly AI assistant. Your team only gets involved for exceptions.

Architecture Overview

Step 1: Create the Application

Create application

Create a new application called “Customer Service”.

Create workflow

Create a workflow called “Returns Handler”.

Step 2: Set Up the Trigger

We’ll use a web widget trigger so customers can initiate returns from your website.

Add Web Widget trigger

Drag a Start node for testing. Later, you’ll replace this with a web widget or email trigger.

Configure test data

Add test data:

{
  "customerEmail": "customer@example.com",
  "customerName": "Sarah"
}

Step 3: Build the Returns Skillset

The skillset handles the conversation: collecting order info, return reason, and item details.

Add Ask a Person node

Name: “Returns Assistant”
Key: returns_conv

Configure the skillset

{
  "agentId": "returns-assistant",
  "name": "Returns Assistant",
  "description": "Help customers initiate product returns",

  "persona": {
    "agentName": "Returns Assistant",
    "organization": "{{store_name}}",
    "tone": "Helpful, empathetic, and efficient. Understand returns can be frustrating. Be apologetic if the product didn't meet expectations but stay positive."
  },

  "inputVariables": [
    {
      "name": "store_name",
      "defaultValue": "Our Store"
    },
    {
      "name": "customer_name",
      "description": "Customer's name"
    },
    {
      "name": "return_policy_days",
      "defaultValue": "30",
      "description": "Number of days for return eligibility"
    }
  ],

  "settings": {
    "maxFieldAttempts": 3,
    "requireConfirmation": true
  },

  "formFieldsMapping": {
    "order_number": {
      "template": {
        "id": "order_number",
        "name": "Order Number",
        "description": "The order number, usually starts with # or found in their confirmation email",
        "type": "text",
        "required": true
      }
    },
    "item_to_return": {
      "template": {
        "id": "item_to_return",
        "name": "Item to Return",
        "description": "Which item(s) from the order they want to return",
        "type": "text",
        "required": true
      }
    },
    "return_reason": {
      "template": {
        "id": "return_reason",
        "name": "Return Reason",
        "description": "Why they want to return the item",
        "type": "selection",
        "options": [
          "Doesn't fit",
          "Not as described",
          "Arrived damaged",
          "Changed my mind",
          "Received wrong item",
          "Quality not as expected",
          "Other"
        ],
        "required": true
      }
    },
    "item_condition": {
      "template": {
        "id": "item_condition",
        "name": "Item Condition",
        "description": "Current condition of the item being returned",
        "type": "selection",
        "options": [
          "Unopened/sealed",
          "Opened but unused",
          "Used once",
          "Used multiple times",
          "Damaged"
        ],
        "required": true
      }
    },
    "additional_details": {
      "template": {
        "id": "additional_details",
        "name": "Additional Details",
        "description": "Any other information about the return, like specific damage or issues",
        "type": "text",
        "required": false
      }
    },
    "preferred_resolution": {
      "template": {
        "id": "preferred_resolution",
        "name": "Preferred Resolution",
        "description": "How they'd like us to resolve this - refund or exchange",
        "type": "selection",
        "options": ["Full refund", "Exchange for different size/color", "Store credit"],
        "required": true
      }
    }
  },

  "knowledgeBaseId": "kb_return_policy"
}

Map input variables

{
  "customer_name": "{{start_1.customerName}}",
  "store_name": "Your Store Name"
}

Sample Conversation

AI: Hi Sarah! I’m here to help with your return. I’m sorry the product didn’t work out. Can you tell me your order number? You’ll find it in your confirmation email — it usually starts with #. Customer: It’s #12345 AI: Got it, order #12345. Which item from that order would you like to return? Customer: The blue sweater AI: The blue sweater — no problem. What’s the reason for the return? Customer: It doesn’t fit right AI: I understand — fit issues are frustrating! What condition is the sweater in? Customer: I tried it on once but it still has tags AI: Would you prefer a full refund, an exchange for a different size, or store credit? Customer: Exchange for a larger size AI: Perfect! Let me confirm: You’d like to exchange the blue sweater from order #12345 for a larger size because it doesn’t fit. The item is opened but unused. Is that correct? Customer: Yes AI: Great! I’m processing your exchange now…

Step 4: Look Up the Order

After collecting the order number, verify it exists and get order details.

Add App Action node

After the Ask a Person event handle, add:

App: Shopify
Action: Get Order

Configure the lookup

{
  "orderNumber": "{{returns_conv.collectedData.order_number}}"
}

Add error handling

Add a Condition node to check if the order was found:

{{shopify_order.success}} == true

If not found, send a reply asking them to double-check the order number.

Step 5: Validate Return Eligibility

Use AI to check if the return meets your policy.

Add Ask AI node

Name: “Validate Return”
Key: return_validation

Configure the validation prompt

You are a returns policy validator. Determine if this return request is eligible.

RETURN POLICY:
- Returns accepted within 30 days of delivery
- Items must be in original condition with tags (except for defective items)
- Sale items are final sale (no returns)
- Swimwear and underwear cannot be returned if opened
- Damaged items can be returned regardless of condition

ORDER DETAILS:
- Order Date: {{shopify_order.response.created_at}}
- Delivery Date: {{shopify_order.response.fulfilled_at}}
- Item: {{returns_conv.collectedData.item_to_return}}
- Item Category: {{shopify_order.response.line_items[0].product_type}}
- Was on Sale: {{shopify_order.response.line_items[0].discount_allocations.length > 0}}

RETURN REQUEST:
- Reason: {{returns_conv.collectedData.return_reason}}
- Item Condition: {{returns_conv.collectedData.item_condition}}
- Preferred Resolution: {{returns_conv.collectedData.preferred_resolution}}
- Days Since Delivery: <calculate from dates>

Respond in JSON:
{
  "eligible": true/false,
  "reason": "Brief explanation",
  "exceptions": ["any policy exceptions that apply"],
  "suggested_resolution": "What we should offer",
  "requires_human_review": true/false
}

Step 6: Route by Eligibility

Add Condition node

Check eligibility:

{{return_validation.response.eligible}} == true

Create two branches

Eligible → Process the return
Not Eligible → Explain and offer alternatives

Step 7: Process Eligible Returns

For eligible returns, generate a shipping label and update the order.

Generate Return Label

Add HTTP Request node

If you use a shipping API (ShipStation, EasyPost, etc.):

Method: POST
URL: https://api.easypost.com/v2/shipments

Configure the request

{
  "shipment": {
    "to_address": {
      "name": "Returns Department",
      "street1": "123 Warehouse St",
      "city": "Your City",
      "state": "ST",
      "zip": "12345"
    },
    "from_address": {
      "name": "{{start_1.customerName}}",
      "street1": "{{shopify_order.response.shipping_address.address1}}",
      "city": "{{shopify_order.response.shipping_address.city}}",
      "state": "{{shopify_order.response.shipping_address.province_code}}",
      "zip": "{{shopify_order.response.shipping_address.zip}}"
    },
    "parcel": {
      "weight": 16
    }
  }
}

Update Order in Shopify

Add App Action node

App: Shopify
Action: Add Order Note or Add Order Tag

Configure the update

{
  "orderId": "{{shopify_order.response.id}}",
  "note": "Return initiated: {{returns_conv.collectedData.item_to_return}} - {{returns_conv.collectedData.return_reason}}. Label: {{shipping_label.response.tracking_code}}",
  "tags": ["return-initiated", "{{returns_conv.collectedData.preferred_resolution}}"]
}

Send Confirmation to Customer

Add Send Reply node

Great news! Your return has been approved. Here's what happens next:

📦 **Return Label**
I've emailed your prepaid return label to {{start_1.customerEmail}}.

📋 **Instructions**
1. Pack the blue sweater securely
2. Attach the return label to the outside
3. Drop off at any UPS location

🔄 **Your Exchange**
Once we receive your return (usually 5-7 business days), we'll ship your new size right away.

**Tracking:** {{shipping_label.response.tracking_code}}

Anything else I can help with?

Notify Fulfillment Team

Add App Action (Slack)

📦 *Return Incoming*

Order: #{{returns_conv.collectedData.order_number}}
Customer: {{start_1.customerName}}
Item: {{returns_conv.collectedData.item_to_return}}
Reason: {{returns_conv.collectedData.return_reason}}
Resolution: {{returns_conv.collectedData.preferred_resolution}}
Tracking: {{shipping_label.response.tracking_code}}

Expected arrival: ~5-7 days

Step 8: Handle Ineligible Returns

For returns that don’t meet policy, explain why and offer alternatives.

Check if human review needed

Add a condition:

{{return_validation.response.requires_human_review}} == true

For human review

Transfer to human:

I want to make sure you get the best resolution. Let me connect you with our customer service team who can review your specific situation. They'll reach out within 24 hours.

In the meantime, is there anything else I can help with?

Then notify your CS team via Slack.

For clear policy violations

Send explanation:

I'm sorry, but I'm not able to process this return. {{return_validation.response.reason}}

Here are some alternatives:
{{#if return_validation.response.suggested_resolution}}
- {{return_validation.response.suggested_resolution}}
{{/if}}
- Contact our customer service team for special consideration
- Check if the item qualifies for our warranty program

Is there anything else I can help with?

Complete Workflow

Adding Email Support

Let customers initiate returns by email too.

Add parallel trigger

Create another workflow with an App Trigger (Gmail):

Filter: Subject contains “return” or “exchange”

Extract customer info

Use Ask AI to extract order number and intent from the email body.

Route to same skillset

Pass extracted data to the same returns skillset, but via email channel.

Testing Scenarios

Test these scenarios to ensure robustness:

Scenario	Expected Behavior
Valid return within 30 days	Approved, label generated
Return after 30 days	Rejected with explanation
Damaged item (any age)	Approved under damage exception
Sale item	Rejected - final sale
Wrong order number	Asked to verify
Customer wants exchange + refund	Clarify which one
Unclear item description	AI asks for clarification

Metrics to Track

Metric	Target	Why It Matters
Self-service completion rate	>80%	Fewer tickets for your team
Average conversation length	<5 minutes	Customer satisfaction
Return approval rate	Monitor	Policy effectiveness
Human escalation rate	<10%	AI handling most cases
Customer satisfaction (post-return)	>4/5	Overall experience

Customization Ideas

Add Photo Upload

For damaged items, let customers upload photos:

{
  "id": "damage_photo",
  "name": "Damage Photo",
  "description": "Photo showing the damage for our records",
  "type": "file",
  "required": true,
  "condition": "return_reason == 'Arrived damaged'"
}

Multi-Item Returns

Handle returns with multiple items by making item_to_return a loop:

{
  "id": "items_to_return",
  "name": "Items to Return",
  "description": "All items they want to return from this order",
  "type": "array",
  "required": true
}

Proactive Return Offers

For very dissatisfied customers, proactively offer returns before they ask.

Troubleshooting

Order lookup fails

Check if the order number format matches what Shopify expects. Some stores use prefixes like “ORD-” that customers might not include.

Policy validation is too strict/lenient

Adjust the AI prompt with more specific rules. Be explicit about edge cases and exceptions.

Shipping label API errors

Verify address formatting. EasyPost and ShipStation have specific requirements for address fields.

Customers abandon mid-conversation

Consider sending an email follow-up if they don’t complete within 30 minutes. Add a timeout handler.

Next Steps

Lead Qualification

See how skillsets work for sales use cases.

Integrations

Connect to more e-commerce platforms.

Knowledge Base

Add your return policy to the knowledge base.

Channels

Enable more communication channels.

Guides

​Prerequisites

​The Business Problem

​Architecture Overview

​Step 1: Create the Application

​Step 2: Set Up the Trigger

​Step 3: Build the Returns Skillset

​Sample Conversation

​Step 4: Look Up the Order

​Step 5: Validate Return Eligibility

​Step 6: Route by Eligibility

​Step 7: Process Eligible Returns

​Generate Return Label

​Update Order in Shopify

​Send Confirmation to Customer

​Notify Fulfillment Team

​Step 8: Handle Ineligible Returns

​Complete Workflow

​Adding Email Support

​Testing Scenarios

​Metrics to Track

​Customization Ideas

​Add Photo Upload

​Multi-Item Returns

​Proactive Return Offers

​Troubleshooting

​Next Steps