Ecommerce & Online Retail

Ecommerce & Online Retail/ Returns & exchanges

Make the next step depend on the item, reason, and order

A return is not one decision. Whether a customer should receive an exchange, refund, replacement, store credit, warranty path, or manual review depends on what was purchased, why it is being returned, when it arrived, and what condition it is in. This use case helps turn that complexity into a clear resolution path.

RETURNS DESK
ILLUSTRATIVE WORKFLOW
CASE CONTEXTTHESE DETAILS SHAPE THE PATH
01The item
02The timing
03The condition
04The reason
RESOLUTION PATHEvaluate the case before choosing the next step
ExchangeRefundReplacementStore creditWarrantyManual review

Doc Search and Data Search provide the relevant product and policy rules.

Eligibility is the intersection of four things.

The AI agent uses product and order details together rather than quoting a generic returns page.

01

The item

Different product categories may have different restrictions.

02

The timing

Return windows may start from purchase, shipment, or delivery.

03

The condition

Opened, used, worn, damaged, or incomplete products may follow different rules.

04

The reason

Changed mind, wrong size, damaged delivery, incorrect item, defect, and missing parts are not the same case.

Resolution should change with the reason

The question is which resolution applies to this situation.

The important question is not simply “Can this be returned?”

CUSTOMER SITUATIONLIKELY PATH TO EVALUATE
  1. 01Wrong sizeExchange availability
  2. 02Changed mindStandard return eligibility
  3. 03Damaged on arrivalReplacement or refund process
  4. 04Wrong item receivedCorrection and return path
  5. 05Missing componentReplacement part or full return
  6. 06Defective productWarranty or defect process
  7. 07Final-sale itemRestriction or exception review
  8. 08Gift purchaseGift-specific refund or credit rules

It is “Which resolution applies to this situation?”

EXCHANGE CHECK

Exchange logic is partly an inventory problem.

A size exchange may be allowed by policy but impossible if the requested variant is sold out.

The AI agent can use current product or inventory data to check:

  • replacement size
  • color
  • model
  • price difference
  • whether the same item is still available

This prevents the customer from entering an exchange flow that cannot be completed. When another variant is available, the agent can explain the practical alternatives.

RETURN WINDOW

Deadlines should be calculated from the correct event.

Return timing can be more complex than “30 days.”

0130 days from delivery
02holiday extensions
03preorder timing
04different windows for members
05shorter periods for certain categories
06separate warranty timelines

When the relevant order dates are available, the AI agent can determine the actual deadline and explain it directly. That is more useful than asking the customer to calculate it from a policy page.

Standard return intake

A compact intake should collect only what the decision needs.

For a standard return, the conversation may only need these details. For a damaged or incorrect item, more information may be needed.

The AI Form node can collect structured details when the decision requires several inputs. The flow should get longer only when the case is more complex.

STANDARD RETURN5 DETAILS
  1. 01order
  2. 02item
  3. 03reason
  4. 04condition
  5. 05desired resolution
MORE COMPLEX CASECollect only the additional information needed

Warranty · replacement · troubleshooting

Some returns should become another workflow entirely.

A customer may think they need a return when the better path is one of these:

WHAT THE CUSTOMER ASKS FORReturn this item
01warranty support
02replacement part
03troubleshooting
04missing-item resolution
05delivery damage claim
06size exchange

That distinction can protect both customer experience and operational cost.

The AI agent should recognize when “return” is only the customer’s starting language, not necessarily the correct business process.

WHEN A PERSON SHOULD DECIDE

Exceptions need a boundary

The AI agent should explain the standard rule without pretending to approve an exception.

A customer may be dealing with any of these situations:

  • one day outside the return window
  • missing packaging
  • returning a restricted item
  • asking for special consideration
  • unusual delivery circumstances

Human Handoff can move the case to a person when discretion is required.

HANDOFF CONTEXTREADY FOR REVIEW

The handoff should include:

OrderItemPolicyReason for the exception request
Human Handoff

A fair path can also avoid an unnecessary full refund

The best resolution is not always the most expensive one.

A missing component may be solved with a replacement part. A wrong size may be better handled with an exchange. A defective product may belong in a warranty process. A damaged shipment may require a carrier or replacement workflow.

Good return guidance helps the business avoid turning every problem into a full refund while still giving the customer a fair path forward.

Doc Search · Data Search · AI Form · Human Handoff

The AI agent uses them together rather than quoting a generic returns page.

01

Doc Search

Find the relevant return, warranty, and product-specific policy rules.

02

Data Search

Compare order details, product restrictions, dates, and connected inventory information.

03

AI Form

Collect structured return details when the case requires several inputs.

04

Human Handoff

Move exception requests to a person with the order, item, policy, and reason attached.

Practical questions

Return eligibility & resolution paths.

Can the AI agent determine return eligibility?

Yes. It can compare order details with the relevant return rules from Doc Search or Data Search.

Can it handle different rules by product type?

Yes. Product-specific restrictions and policy conditions can be applied to the item being returned.

Can it check whether an exchange variant is available?

Yes, when current inventory information is connected.

Can it collect return details?

Yes. The AI Form node can collect structured information when the case requires it.

What happens when the customer asks for an exception?

The agent can explain the standard policy and move the case to Human Handoff when staff judgment is required.

What the customer should leave with.

By the end of the conversation, the customer should understand:

  • whether the item appears eligible
  • which rule applies
  • what resolution options are available
  • whether stock affects an exchange
  • the deadline
  • what information or item condition is required
  • whether staff review is needed

That clarity is the real value of the use case.