Skip to content
ToolDeck

Ecommerce

Support Ticket Triage

Classifies every store message and drafts answers for the questions you answer on repeat.

Saves about ~5 hrs per week

What it does

  • Classifies incoming support messages by topic
  • Drafts answers for order status, sizing, and returns
  • Routes uncertain tickets to you with context
  • Reports weekly on ticket volume and deflection

How it works

  1. Support message arrives
  2. AI classifies the intent
  3. Common topics get a drafted reply
  4. Uncertain ones route to the owner

What's included

  • Helpdesk connection
  • Answer bank
  • Triage + routing rules

Set it up yourself

Follow these 4 steps in order. You will need accounts for: Shopify.

Step 1: Connect your helpdesk

  1. Choose the source: Gorgias, Zendesk, or your shared inbox via IMAP.
  2. Create the API key for the helpdesk (or generate an IMAP app password).
  3. In the workflow, add the trigger 'new ticket/message' with the key.
  4. Capture: message text, customer email, order ref if present.
  5. Test a real ticket comes through.

Watch out: Include the order reference where available — the AI's answer quality jumps when it sees the actual order.

Step 2: Set the answer bank

  1. Write approved answers for your top question types (order status, sizing, returns, shipping, product question).
  2. Save them in a doc or sheet 'Answer Bank' with a topic label.
  3. Give the AI the answer bank as reference so replies stay on-brand.
  4. Add a rule: never quote prices or promises not in the bank.
  5. Test the bank retrieval.

Watch out: The answer bank is your control — the AI should paraphrase the bank, not improvise policies.

Step 3: Add the classifier

  1. Build: 'Classify this support message into one of: {topics}, or unknown. Return JSON: {topic, draftReply, confidence}.'
  2. For known topics with confidence > 0.8, attach the drafted reply.
  3. For unknown or low confidence, mark 'route to human'.
  4. Paste with API key.
  5. Test on 20 past tickets and eyeball the classification.

Watch out: Unknown topics going to a human is the safety net — an over-confident classifier answering wrong is worse than no automation.

Step 4: Set routing rules

  1. Define the decision: known+high-confidence → draft to queue; unknown → human.
  2. For the human route, prepend the AI classification as context ('AI thinks: sizing question — please check').
  3. Optional: auto-send the drafted reply for the safest topics (order status) after a week of approvals.
  4. Log everything (ticket, topic, sent/queued) to a sheet.
  5. Review the weekly report and tighten the answer bank.

Watch out: Start draft-only for ALL topics; graduate only the safest to auto-send.

Frequently asked questions

What does the Support Ticket Triage automation do?

Classifies every store message and drafts answers for the questions you answer on repeat. Classifies incoming support messages by topic

How much time does Support Ticket Triage save?

It is estimated to save ~5 hrs per week. Actual savings depend on your volume.

Which tools does Support Ticket Triage use?

It works with Shopify.

What does Support Ticket Triage include?

Helpdesk connection, Answer bank, Triage + routing rules.

Can I set up Support Ticket Triage myself?

Yes. The do-it-yourself guide on this page walks through 4 steps, from getting access to testing the finished automation.