Platform: All platforms | Plan: Premium and Enterprise | Availability: US merchants only.
Overview
The RMA Review Agent is an AI-powered review layer for Returns > Return policy > Fraud prevention. For every return request (RMA), it assesses fraud risk, explains that risk in plain language, and recommends a next action for low-risk returns so you can process them faster.
The RMA Review Agent is an upgrade to the existing Fraud prevention feature. Your anti-fraud workflows, auto-flag settings, and return history display all stay exactly as they are, so you can keep defining your own customized rules the same way you do today.
What changes is the risk assessment itself. If your organization is enrolled, the original risk assessment is upgraded to the AI-powered RMA Review Agent. If you're not enrolled, you'll keep using the original Fraud prevention feature without any change at all.
Why use the RMA Review Agent
Reviewing every return by hand takes time, and the information you need to make a decision is often spread across several pages. The RMA Review Agent helps you:
Benefit | What it does |
Reduce fraud loss | Flags high-risk returns earlier and more consistently, so fewer risky requests get approved by mistake |
Improve the shopper experience | Recommends earlier refunds or exchanges on returns it identifies as safe |
Improve operating efficiency | Gives clear guidance on low-risk returns, so your team can focus on the returns that truly need a closer look |
Enable the RMA Review Agent
Go to Return policy > Fraud prevention in your AfterShip Returns account.
Under Risk detection and prevention, turn on the Risk assessment toggle.
Optionally, select Auto-flag high-risk returns for manual review. When this is on, any return the agent scores as high risk is automatically flagged, and its automation rules are suspended until someone reviews it manually.
Your existing anti-fraud workflows and return history settings keep running as before. Therefore, no changes are needed there.
Note: You can turn off Risk Assessment at any time in Fraud prevention settings without affecting your other automation rules.
How the risk assessment works
AfterShip's internal Fraud Risk Model automatically scores every RMA. You don't need to trigger anything. The model reviews:
Photo evidence, when a shopper uploads photos. It checks whether the photo looks genuine, whether it matches the ordered product, and whether it matches the stated return reason.
Risk signals, such as shopper return history, order patterns, logistics details, and cross-merchant data. The model weighs combinations of signals rather than reacting to any single one.
Based on this analysis, the agent assigns one of three risk levels:
1. Low risk
What it means: Fraud is unlikely based on the shopper’s history and return details.
What you should do: Review the recommended action and process the RMA.
2. Medium risk
What it means: Some signals warrant a closer look.
What you should do: Review the return manually before approving it.
3. High risk
What it means: There are clear signs of potential fraud or policy abuse.
What you should do: Review the return manually and conduct a closer inspection.
View the risk assessment card
Open any return request to see the AI risk assessment card on the RMA detail page. The card includes:
The risk level (low, medium, or high)
The risk signals that were triggered
A plain-language summary covering the current RMA, the shopper's purchase and loyalty history, and the shopper's return behavior in your store
Report fraud and Mark as safe buttons, so you can give feedback that helps train the model
For low-risk returns, a recommended action, such as Approve > Refund (when in transit)
For medium- and high-risk returns, the card shows the risk level and signals only. It doesn't show a recommended action; these returns require manual review.
Apply a recommended action
For low-risk returns, the agent suggests a sequence such as Approve > Refund or Approve > Create exchange. You can:
Accept the recommendation to let the agent carry it out
Edit the refund or exchange timing before accepting
Setting | Available timing options |
Refund timing | When in transit (default), when delivered, or when marked as received. |
Exchange timing | When the RMA is approved, when in transit, when delivered, or when marked as received. |
Green returns (where the shopper keeps the item and no return shipment is needed) resolve immediately after you approve them, and their timing can't be edited.
Tip: If the recommended timing is earlier than your current automation settings, the card lets you know. This means the agent has identified an opportunity to give the shopper a faster refund or exchange without added risk.
The first time you apply a recommendation, you can choose whether to keep a confirmation step for future low-risk returns, or apply recommendations automatically going forward.
Note: The RMA Review Agent never takes action on its own. It only refunds, exchanges, or resolves a return after you select Accept.
Track agent activity
Every risk assessment and action the agent takes is logged in the return request's timeline. You'll see entries such as:
Return risk level assessed as High via AI agent
Recommended actions applied by [User] via AI agent
These entries stay in sync with the rest of your return request activity, ensuring a complete audit trail.
Give feedback to improve accuracy
You can help improve the Fraud Risk Model's accuracy over time by using the feedback buttons on the risk assessment card:
Select Report fraud if you believe a return is riskier than the agent indicates, even on a low-risk card.
Select Mark as safe if you believe a medium- or high-risk return is actually safe.
Your feedback corrects the current case and helps train future versions of the model.
FAQs
1. Does the RMA Review Agent automatically approve or refund returns?
No. In the current version, you must select Accept for any action to run. Automatically flagging high-risk returns for manual review is a separate, optional setting.
2. Can I disagree with the agent's risk assessment?
Yes. Use Report fraud or Mark as safe on the risk assessment card to correct the current case and help improve future assessments.
3. Will this conflict with my existing automation rules?
No. Your existing automations and anti-fraud workflows continue to work as before. If a recommended timing is earlier than your current configuration, the card calls this out so you can decide whether to apply it.
4. Is the RMA Review Agent available outside the US?
Not yet. Due to data and privacy considerations, the RMA Review Agent is currently available to US merchants only.
5. What if I don't want to use the RMA Review Agent?
If your organization isn't enrolled, you'll continue using the original Fraud prevention feature with no change to your workflow.
Key takeaways
The RMA Review Agent assesses every return request for fraud risk and shows the result on an AI risk assessment card.
Low-risk returns get a recommended action you can accept or edit; medium- and high-risk returns are flagged for manual review.
The agent never takes action without your approval.
All predictions and actions are logged in the return request timeline.
The feature is currently available to Premium and Enterprise merchants in the US only.