Playbook

Support automation for Shopify stores: when AI saves you 30 hours a week (and when it doesn't)

A pragmatic playbook for automating Shopify support — macros, AI agents, and the math on when each technique pays back. Real numbers from 100-2000 ticket/month stores.

By Botapolis editorial2026-05-148 min read
Steps5

This article contains affiliate links. Pricing and terms are set by the vendor. details

Support is the silent margin-killer of growing Shopify stores. The first year of a $30k MRR store, the founder answers tickets in their inbox between order-fulfilment runs. By year two with three part-time agents handling 800 tickets/month, support is suddenly a $4k/mo line item — and most of those tickets are versions of the same 5 questions.

This guide is the framework we use to automate that volume without breaking the customer experience. Real numbers from running these flows on 100-2000 ticket/month stores over 18 months.

1. Tag first, automate second

Every automation playbook starts with the same step everyone wants to skip: tag the last 200 tickets and find the actual patterns.

Open the tickets in your helpdesk (Gorgias, Tidio, Re:amaze, Shopify Inbox — doesn’t matter) and assign each one a category. We use these:

  • order_status — where is my package, tracking issues, ETA questions
  • refund_request — return for refund, order didn’t arrive, defect claims
  • address_change — update shipping after purchase
  • sizing_question — pre-sale size / fit questions
  • shipping_policy — pre-sale shipping cost / time / country
  • product_question — pre-sale spec / compatibility / how-to questions
  • complaint — customer-anger tickets (regardless of underlying issue)
  • cancellation — cancel order before fulfilment
  • vip — repeat customer, big order, or social-flagged
  • other — everything else

On every store we’ve tagged, the top 5-7 categories cover 75-85% of total volume. The “other” bucket is where automation can’t help — those need humans.

2. Live-data macros for the top 3 categories

The first automation layer is macros that fetch live data and insert it into the reply. This is plain-helpdesk territory — Gorgias does it best, Tidio does it via fetch button, Re:amaze does it manually. Whatever tool you’re on, set up these three first:

Order status macro

Pulls: customer’s most recent order, carrier name, tracking number, tracking URL, expected delivery date.

Reply template:

Hi {first_name},

Your order #{order_number} shipped via {carrier} on {ship_date} and is on the way!
Tracking: {tracking_url}
Expected delivery: {expected_delivery_date}

Let me know if you have questions.

One-click sends. Time per ticket: ~15 seconds (vs ~5 minutes manual: open Shopify, find order, copy tracking, paste, send).

Refund request macro

Pulls: order details, refund-eligible status (based on return-window policy), refund amount.

Two variants:

  1. In-policy (within 30 days, item not on final-sale): macro processes refund via Shopify, replies with confirmation.
  2. Out-of-policy: macro replies with policy details and escalates to human review.

Time per ticket (in-policy): ~30 seconds with refund. Out-of-policy reverts to human handling.

Address change macro

Pulls: order details, current shipping address, fulfilment status.

Branch logic:

  • If fulfilment hasn’t started → macro updates the shipping address in Shopify, replies with confirmation
  • If shipping label printed but not handed to carrier → macro shows agent a “contact warehouse” prompt
  • If carrier has the package → macro replies with carrier-side change instructions (usually impossible at that point)

Time savings per ticket: ~3 minutes (manual change is fiddly).

These three macros alone, on the 612-ticket month we tracked for our Gorgias review, saved an estimated 65 agent-hours.

3. AI deflection for the FAQ tier

The second automation layer is generative AI for stable-answer questions. The right targets:

  • Sizing questions (answer lives in your size chart + product copy)
  • Shipping policy (answer is a small set of facts about your shipping zones / times / costs)
  • Return policy (answer is in your return-policy doc)
  • Product spec questions (answer is on the PDP)

These work for AI because the answer is stable and discoverable — the model doesn’t have to invent anything, it just retrieves and rephrases.

The bad targets for AI:

  • Refund-eligibility edge cases (need human judgement)
  • Complaint resolution (need empathy)
  • Custom requests (need creativity)

Implementation depends on your tool:

  • Tidio Lyro (review) — best at this for sub-$50k MRR stores. Per-conversation pricing.
  • Gorgias Automate — purpose-built, included in Pro+ tiers. Better at order-context queries than Tidio.
  • Re:amaze Helix — newer, weaker than the above two as of 2026.

Train the AI on:

  • Your full product catalogue (via Shopify integration)
  • Your published policies (shipping, returns, FAQ pages)
  • 30-50 hand-crafted Q&A pairs for the questions your tagging exercise surfaced

Then audit-check 20-30 AI replies in the first week. The model will hallucinate occasionally — usually on spec questions where the underlying PDP copy is ambiguous. Fix the PDP, not the AI.

4. Routing for escalations

Some tickets MUST reach a human within minutes — automation that delays them hurts more than it helps. Build routing rules for:

  • VIP customers (top 10% by CLV) — auto-assign to senior agent, skip the macro / AI layer entirely
  • Complaint tickets (negative sentiment in subject or body) — escalate immediately, never AI-deflect
  • Fraud flags (high-risk score from Shopify) — go to founder or senior support, not standard queue
  • Repeat-contact within 24h — if a customer messages twice within a day, the first reply didn’t satisfy. Send to a human, not back through the loop

Most helpdesks let you express these as automation rules. Set them up at the SAME TIME you set up macros and AI deflection — not after the fact when an angry customer escalates publicly because the AI told them to “check the FAQ”.

5. Measuring what matters

Volume deflection is the obvious metric — “the AI handled 40% of tickets, the macros handled another 30%, agents handled 30%”. But pure deflection can hide damage.

Track three numbers:

  • Deflection rate — % of tickets resolved without human touch
  • CSAT after AI/macro touch — has automated resolution dropped satisfaction?
  • Reopen rate — what % of “resolved” tickets come back within 7 days?

A flow that deflects 60% of tickets but reopens 25% of them is actually creating MORE work (you handled the issue once, the customer is now annoyed and re-asking). Deflection that holds reopen-rate steady IS a win.

Our threshold rule: if a deflection bucket’s reopen rate exceeds 12%, the rule comes off and goes back to manual. AI / macros have to MATCH human-quality on closure, not just match human-volume.

6. The math on payback

Concrete from our 600-700 ticket/month test store:

LayerTickets deflected/handledAgent time savedCost
Live-data macros (3 categories)~280/mo~65 hours/mo$0 (free in Gorgias)
AI deflection (FAQ tier)~80/mo~12 hours/mo$0 (Pro tier includes Automate)
Routing rules(none directly deflected)~5 hours/mo (fewer escalation misfires)$0
Total~360/mo deflected~82 hours/mo saved~$0 incremental

At $25/hr part-time agent loaded cost, that’s $2,050/month in labor saved. The helpdesk Pro tier ($360/mo) pays back nearly 6×.

The catch: this math is for a 600-ticket store with a tagged automation strategy. A 100-ticket store hits the same percentage gains in raw numbers, but the absolute savings (~$340/mo labor) are smaller and the helpdesk Pro tier becomes harder to justify. Below 200 tickets/month, run macros only and skip the AI tier.

Gorgias

Helpdesk built for Shopify — turn support tickets into revenue.

from $10/mo

Try Gorgias free

Where to start by store size

  • Sub-100 tickets/month: Tidio free tier + Lyro AI (~$40/mo) + 3 simple macros. Total cost ~$40/mo, saves maybe 10-15 hours.
  • 100-500 tickets/month: Tidio paid + Lyro, OR Gorgias Basic. Around $90-$140/mo + AI add-on. Significant payback.
  • 500-2000 tickets/month: Gorgias Pro is the right answer. The macro engine, AI deflection, and routing rules collectively save 50-100 hours/month at this volume.
  • 2000+ tickets/month: Enterprise tier conversation with Gorgias or Kustomer. Pricing custom; specialised CSM helps.

What this looks like as a four-week ramp

  • Week 1: tag 200 tickets, identify top 5 categories. Don’t automate yet.
  • Week 2: build 3 live-data macros for the top order-context categories. Measure time-per-ticket before / after.
  • Week 3: enable AI deflection on FAQ-tier categories with knowledge base trained from existing docs. Audit-check 30 responses.
  • Week 4: set up escalation routing. Measure CSAT and reopen rate. Tune rules where reopen exceeds 12%.

By end of month one, support cost should drop 30-50% vs baseline — not from layoffs, but from agents covering the same volume in half the time. That’s the win.

Last updated: May 2026.

On this page