WhatsApp AI Agent for Returns

Go to the AiSensy Signup Page. Create your account using either:

• Your Google account, or

• Your business email address by filling out the sign-up form.

Once signed up, Log in to AiSensy using the same Google account or email credentials.

Once you are logged in, navigate to and click on AI Agent available on the navbar on the left.

Fill in the details given below:

• Website (mandatory)

• Business type

• Anything special it should know? (optional)

• Business context (optional)

Once you fill these details correctly, proceed to hit '✨ Create agent' button. AiSensy automatically builds an agent trained on your business.

When you add your website, AiSensy AI engine auto-generates your business description. Set your agent's name, greeting, and a few ground rules it must always follow.

Fill the details given below:

• Website (fetched automatically)

• Agent/business name

• Greeting message (optional)

• Store currency

• Business type

• What the business does

• Ground rules

In this step, you have to tell the agent about your business website. You can include return windows, excluded products, item-condition requirements, return reasons accepted, pickup charges, packaging instructions, and refund timelines.

You can also train your agent to explain when the return window starts and when the refund timeline begins.

You can submit this information in the form of PDF, DOC, TXT, or CSV files. The agent entirely relies on the kind of information you submit. Hence it always responds according to it and never gives generic replies.

Every source shows "Ready" once learned.

Perform the below mentioned steps:

• Crawl a page (website URL)

• Upload files (PDF, DOC, TXT, CSV - up to 50MB each)

• Paste text (policies, scripts, price lists)

This step is entirely optional. You can add packing instructions or pickup preparation guides in this step. This step usually controls the images your agent sends. It doesn't configure collecting damaged photos from customers.

When you add your website, the WhatsApp AI Agent automatically fetches the image from it. You can control the visuals the agent can share inside the WhatsApp chat.

Untick any image it shouldn't send.

You can:

• Create Image library (enable/disable per image)

• Search images

Open Add Skills and check the prebuilt list first - if the skill you need isn't there, hit Create custom, name it Returns & Exchange, and pick Task skill as the type.

• What kind of skill is this? (Task skill)

• Icon & name: Returns & Exchange

• When should the agent use it?

• Instructions

• Add skills

Now, you must make sure that the instructions fed to the agent are clear and easy to understand. The agent behaves according to the instructions given to it.

One such template example is:

“When a customer says ‘I want to return my order’, ask for their order ID, the item they want to return, and the reason, one question at a time. Follow the required verification steps and check eligibility using the store’s return policy and order details. If eligible, guide them through submitting a return request. Explain pickup and refund timelines only from the information provided. Refer exceptions to the support team, and never confirm a return or refund unless the connected system confirms it.”

Here, you can add your return window, item condition requirements, non-returnable products, and any return charges to the agent’s knowledge.

The agent uses your instructions and connected catalog to help customers find and buy their desired products.

For more info, tap here to learn how to create a custom skill

The skill you just created lands in the Skills list. You can now switch it on, expand it to set your qualifying questions, your cut-off and where a qualified applicant should go, then hit Done.

All you have to go through is:

• Skills list

• Enable the returns skill

• Expand to configure its variables, templates & ads

• Skills enabled counter

In this step, you can train your agent to check the request against your policy and order records, explain the available options, and obtain confirmation before submitting a return.

The agent will then be able to distinguish a submitted request from an approved return or completed refund.

In this step, you can train your agent to use triggers for policy exceptions, disputed eligibility, missed pickups, overdue refunds, failed return submissions, or requests for a support representative.

You can decide as to what happens when the AI agent can't help. You can write the fallback message, switch on Hand off to a human, and add the phrases that should pull a person into the WhatsApp chat. 

What you must fill:

• Fallback message

• Where your agent runs: up to 3 test numbers

• Hand off to a human: toggle

• Trigger phrases

• Handoff message

It is one of the most crucial steps while building a returns WhatsApp AI agent.

You will need the agent to retrieve order details, check eligibility against purchase records, create return requests, or retrieve pickup and refund status. 

You can skip this step if the agent only explains policies and shares a return portal link.

Actions let your agent fetch live information or complete tasks using connected tools and APIs - checking an order status, creating a CRM lead or booking an appointment.

Hit Add custom API action, tell it when to call it, then paste your cURL - the request fills itself in.

• Action name

• When should the agent call it?

• Paste cURL

• Request: method + endpoint URL

• Headers

• Params

• Body

• Auth

• Test call

Click on - Next: variables

Before making the agent go live, you can test it. You can test eligible and ineligible returns, partial returns, duplicate requests, damaged items, and integration failures.

Please ensure that the agent doesn’t promise approval, pickup dates, or refunds without confirmation from the relevant system.

You can chat with your agent before your customers do with AiSensy's 500 free AI messages for a trial.

Publish, switch the AI on & you're live.

• Test your Agent panel (Train/Live modes)

• Publish

• AI on/off toggle