AI Actions: Setting up your AIVA to handle Shopify order status calls

This tutorial explains how to configure an AI Voice Agent to automatically handle “Where is my order?” calls using the Shopify integration and the Get Order Details AI action. By the end, your agent will be able to identify callers, retrieve their order in Shopify, and respond conversationally based on payment and fulfillment status, without transferring to a human agent.

What you will build

A caller dials your Aircall number. The AI Voice Agent answers, asks for the order number and a verification detail such as email or phone, queries Shopify in real time, and responds בהתאם to the order status.

Example conversation:

Caller: “Hi, I ordered something last week and I haven’t heard anything since. Where is it?”
Agent: “Happy to help. Could you give me your order number? It usually starts with a # symbol and is six digits long.”
Caller: “#12345.”
Agent: “Thanks. To make sure I’m speaking to the right person, could you confirm the email address on the order?”
Caller: “It’s alex@example.com.”
Agent: “Got it, your order is paid and on its way. It should arrive shortly. The shipment contains two items: a navy hoodie size M and a pair of black socks. Anything else I can help with?”

Step 1: Enable Get Order Details on your Shopify integration

AI actions follow a two step model. First, enable the capability on the integration. Second, assign it to an AI Voice Agent.

Steps:

1. Open the Aircall Dashboard > Integrations & API
2. Select your Shopify installation.
3. Click the AI agent capabilities tab.
4. Toggle Get Order Details on.

Important: Enabling this toggle does not affect live calls or existing configurations. It only makes the action available for assignment to AI Voice Agents. If it is off, the action will not appear in the agent setup.

Step 2: Add the action to your AI Voice Agent

Now assign the action to a specific agent and define how it behaves.

Steps:

1. Go to AI Agents in the left navigation.
2. Select your agent, for example “Shopify Demo”.
3. Scroll to Functional abilities and click AI actions.
4. Click + Add AI action.

5. In the side panel, open the dropdown and select Get Order Details under Shopify.

Note: If Get Order Details is missing, return to Step 1. The toggle may be off or the Shopify integration disconnected.

Note: If you have multiple Shopify installations, each has its own settings. Enabling the action on one does not apply to others.

Step 3: Review the input and output

After selecting the action, review the preset instructions and data contract. This defines what the agent collects and what it can share.

Information collected from the caller

The agent must collect the order number and at least one identifier, email or phone, before performing the lookup. This ensures accurate identification and prevents data exposure.

Information returned to the agent

The action returns a single object called order. This includes all fields the agent can reference in the conversation, such as items, totals, fulfillment, and tracking details.

At this stage, you can already activate the action. The agent will ask for required details, perform the lookup, and respond appropriately. The next step focuses on customization.

Step 4: Tailor behavior with Advanced settings

You can refine how the agent behaves using two methods:

• Custom instructions: Apply only to this action. Ideal for triggers, conditional responses, and failure handling.
• Conversation Guidelines: Apply to the entire call. Best for multi action workflows and transfer logic.

Custom instructions

Open Advanced settings to configure rules. Each line represents a rule the agent evaluates when deciding how to act. Example rules for this use case:

Note:
• Use this when the caller has a problem with his ongoing order or wants to know where his package is.
• If the Financial Status is PAID, reassure that the order will arrive shortly and share the other details.
• If the order status is PENDING, tell the customer that there might be a problem with the used payment method and that he should please check his emails or reach out to the support.

How these rules work

• Line 1 (trigger). Tells the agent when to reach for the action. Without this, the agent decides on its own based on the caller’s phrasing - that usually works, but a clear trigger sentence reduces false positives (e.g. callers asking about returns or refunds).
• Line 2 (PAID branch). When financial_status on the returned order equals PAID, the agent reassures the caller and reads back the relevant detail (line items, tracking, expected delivery).
• Line 3 (PENDING branch). When financial_status is PENDING, the agent flags a likely payment issue and points the caller to email support - instead of falsely promising shipment.

Three patterns work well when you write your own custom instructions: 

• triggers ("Use this when the caller asks about delivery or order status")
• conditional responses ("If financial_status is REFUNDED, confirm the refund and the expected timeline")
•  hand-off rules ("If the order cannot be found after one attempt, transfer to the support branch"). 

Keep each rule on its own line and refer to fields by their names in the order schema. Avoid contradictory rules - the agent will pick one and ignore the other.

Step 5: Save and test

1. Click Add AI action.
2. Confirm the action appears in the enabled list.
3. Use the Test call button to simulate calls.

Test scenarios

• Order with PAID status: Agent reassures and shares details.
• Order with PENDING status: Agent flags payment issues.
• Invalid order number: Agent asks for verification or transfers.

Tailoring this to your business

You can enhance the setup with additional rules:

After each change, perform a new test call. Small wording changes can significantly impact behavior.

What this action can and cannot do