Answer every call on the first ring.
Leads call after 5 PM and hit voicemail.
Est. deploy 2–3 weeksSee the walkthrough
Intake & Scheduling · Ecommerce
Prepare return decisions without guessing policy.
Support searches policies for every return request.
For: Returns managers and the team responsible for intake & scheduling in ecommerce.
01 · The outcome
What good looks like.
The team has a proposed return disposition awaiting human authorization, with source evidence and an explicit approval before any external action. Track time to review a complete return request against a baseline; do not assume automation improves it.
A connected workflow that uses order date, item status, stated issue, and current return policy to map the request to policy clauses and flag ambiguous or high-value cases. It produces a proposed return disposition awaiting human authorization.
Where AI helps, and where a person decides.
AI can map the request to policy clauses and flag ambiguous or high-value cases. It drafts and organizes; it does not authorize spending, commitments, clinical decisions or customer communication. The returns manager verifies the source facts and approves the final action.
Returns manager is accountable. People approve customer-facing, sensitive, financial, and legally binding actions.
- Returns manager: Capture the source and scope
- Returns manager: Review the proposed outcome
- Returns manager: Release only the approved version
- Returns manager: Check the result and exceptions
A named person approves customer-facing messages, spending, record changes and binding commitments before release. Pre-approved scripts are versioned; exceptions stop for review.
02 · The handoffs
How the work moves.
Every handoff has an owner. Select a step to open or close its detail.
Start when: A customer opens a return request. Connect only order date, item status, stated issue, and current return policy. The returns manager confirms the request is authorized and in scope.
Ready when: The source reference, permissions and accountable owner are recorded.
Use the authorized records to map the request to policy clauses and flag ambiguous or high-value cases. Preserve the original language alongside the draft and label uncertain fields.
Ready when: Each proposed fact links back to a source, and missing information is visible.
Check a proposed return disposition awaiting human authorization against the original records and current policy. Correct errors and escalate exceptions to the qualified specialist; do not auto-approve.
Ready when: Returns manager has recorded approval or a reason to hold the item.
After explicit approval, save a proposed return disposition awaiting human authorization in the system of record. Send only approved messages to verified recipients. Record the version and prevent duplicate delivery.
Ready when: The approved record is saved; any delivery is confirmed or placed in the retry queue.
Review time to review a complete return request each week, sample completed records for quality, and resolve held items. Change the rules only after reviewing actual errors and staff feedback.
Ready when: Exceptions have owners, the metric has a baseline, and corrective actions are assigned.
Sample record.
$ david run --workflow=commerce-return-triage
ILLUSTRATIVE RECORD / NOT LIVE CUSTOMER DATA
- Example
- Illustrative only — not a real client record
- Reference
- COMMERCE-RETURN-TRIAGE-DEMO-001
- Trigger
- A customer opens a return request.
- Owner
- Returns manager
- Prepared output
- a proposed return disposition awaiting human authorization
- Review status
- Awaiting human approval — no external action taken
- Missing information
- Owner to verify source completeness
- Measure
- Time to review a complete return request
Start with your tools.
Connect Shopify, Gorgias, Klaviyo through approved APIs or controlled exports. Map record identifiers first; test permissions and field updates in a sandbox. Vendor names describe possible connections, not a guarantee of an available integration or license.
When things go wrong.
- A named person approves customer-facing messages, spending, record changes and binding commitments before release. Pre-approved scripts are versioned; exceptions stop for review.
- Use only authorized sources and role-based access. Keep an audit trail of source records, drafts, approvals and delivery status; deduplicate retries.
- If an integration fails, a source is missing or confidence is low, pause the action and assign the original request to the accountable owner.
Try this week / no DAVID required
A useful first step.
This week, have your returns manager time five examples of this work. Use redacted copies of order date, item status, stated issue, and current return policy to make a proposed return disposition awaiting human authorization manually, then list which fields required judgment.
What to measure.
- Time to review a complete return request
- Minutes of human handling per completed item
- Drafts needing substantive correction
- Exceptions unresolved beyond the agreed review window
Estimate your time back.
20.8
$11,232
03 · More workflows to consider
Related workflows.
Route every inquiry, 24/7.
Email, calls and webforms produce disconnected intake trails.
Est. deploy 3–5 weeksSee the walkthrough
Prepare every treatment start before arrival.
Missing forms delay the first treatment visit.
Est. deploy 4–6 weeksSee the walkthrough
Want this running in your business in 2–4 weeks?
Free 30-minute call. We'll tell you honestly whether AI fits, and where.