Recipe — Build a Returns taxonomy with sub-tags

A complete worked example. By the end you’ll have a clean Returns taxonomy with four sub-tags, AI auto-tagging enabled on each, and an automation that escalates defective-product returns.

What you’ll end up with

Returns (Topic — root)
  ├── Defective product
  ├── Wrong item
  ├── Changed mind
  └── Size exchange

Plus one automation:

When a conversation is tagged Defective product, assign to the QA team and add the metadata tag Escalated.

Before you start

• Admin permissions on Tag Management
• A QA team configured in Settings → Teams
• The Metadata tag Escalated already exists, or you’ll create it as part of this recipe

Step 1 — Create the parent

In Settings → Tag Management, create a root tag:

• Name — Returns
• Description — Customer wants to return a product. Mentions returning, refunding, sending back, or no longer wanting an item.
• Type — Topic
• Parent — (empty — this is a root)
• AI Auto-Tag — On

Step 2 — Add the sub-tags via Quick Add

Click Quick Add in Tag Management. Paste:

Returns
  Defective product
  Wrong item
  Changed mind
  Size exchange

Atender creates the four sub-tags as children of the existing Returns parent (matched by name).

Step 3 — Fill in sub-tag descriptions

For each sub-tag, open the edit panel, set the description, and turn on AI Auto-Tag.

• Defective product — Customer received a product that doesn't work, is broken, has manufacturing defects, or is otherwise faulty out of the box.
• Wrong item — Customer received a different product than they ordered. Wrong size, wrong color, wrong model, wrong product entirely.
• Changed mind — Customer no longer wants the product even though it's working as described. Buyer's remorse, found a better option, no longer needs it.
• Size exchange — Customer received the right item but wants a different size. Mentions sizing, fit, too small, too large.

Save each.

Step 4 — Wire the QA-escalation automation

Build a new Automation:

• 1 — Trigger — Tag added
• 2 — Condition — tag equals Defective product
• 3 — Branch — Always
• 4 — Action 1 — Assign conversation → Team: QA
• 5 — Action 2 — Add tag → Escalated (Metadata)
• 6 — Throttle — Per conversation, 1 per 24 hours

The throttle ensures that if the AI re-confirms the tag (which it sometimes does as a conversation evolves), the rule doesn’t re-escalate.

Step 5 — Verify

Send a test inbound email mentioning a defective product:

Hi, the speaker I bought from you last week stopped working. The left channel is dead.

Within a few minutes:

• The auto-tagger should apply Defective product (and probably the parent Returns).
• The automation should fire, reassign to the QA team, and add Escalated.

Open the conversation. You should see all three tags and the assignment to QA in the activity log.

Variants

• Per-region routing. Add a condition on contact.country so only customers in specific regions go to the global QA team; route others to a local one.
• Severity-based escalation. Add a Metadata tag Critical (e.g., for safety-related returns) and have a separate automation that pages the on-call manager when both Defective product and Critical are present.
• Skip the automation, just tag. If you don’t want auto-routing, skip Step 4. The taxonomy alone gives you filterable views and reporting.

Troubleshooting

• Symptom: Auto-tagger applies Returns but not the specific sub-tag. Fix: The sub-tag’s description isn’t specific enough. Compare it against Returns’s description — they should be clearly distinguishable. Add concrete phrasings the customer might use.

• Symptom: Automation fires multiple times for one conversation. Fix: The throttle scope isn’t set. Add per conversation, 1 per 24 hours.