A return form is efficient and slightly tragic. It turns "these are too small" straight into a refund, when what the customer wanted was the next size up. A conversation can do better: confirm the policy, offer the exchange, and only start a return if that's really what the customer wants.
Start with your policy, written clearly
Your agent can only apply a policy that's written down unambiguously. Make sure yours states:
- The return window, and from when it's counted (order or delivery)
- Conditions: unworn, tags on, original packaging
- Exclusions: sale items, personalised goods, hygiene products
- Who pays return shipping
- How long refunds take, and to which payment method
Add it to your knowledge base as its own document.
The exchange-first flow
- Identify the order and item, with verification.
- Check eligibility against the policy and the order date.
- Ask why, briefly: too small, too big, not as expected, faulty.
- Offer the right fix:
- Size issue: the next size, with stock checked
- Colour or style: alternatives in stock
- Faulty: straight to a person
- Confirm the exchange or return explicitly before acting.
- Explain the next step: label, drop-off, timeline.
Build it once as an AI Studio workflow and it runs the same way on your website, WhatsApp and Instagram.
Keep people for the exceptions
Hand off when:
- The item is damaged or faulty
- The request is outside the window but the customer has a reason
- The order value is above your threshold
- The customer is upset
These are the conversations where a person's judgement protects both the relationship and your margin.
Reduce returns before they happen
The best return is one that never starts. Answering fit and spec questions accurately before purchase prevents many of them. That's where an AI agent with your size charts, plus live video for high-value items, pays off twice. See how live video shopping reduces returns.
What to measure
- Return requests that became exchanges
- Time from request to resolution
- Return reasons, by product (useful feedback for your buying and product pages)
- Exceptions handed to people
Treat return reasons as product data. If one line keeps coming back as "too small", the fix might be a size-chart note, not a better returns flow.
