For the 10 to 15 questions that make up most small-store inbox volume, a well-written saved reply is still faster and safer than an AI draft. A macro sends in about 3 seconds, needs no review, and cannot invent a policy. Pretending otherwise is hype. AI wins elsewhere, and this article shows exactly where.
Key takeaways
- Macros beat AI on fixed policy text, legal wording, and cutoff notices.
- AI drafts beat macros on order-specific, multi-question, and non-English emails.
- Run one source of truth so you maintain each answer once, not twice.
- Test the split on 25 real tickets before committing to a setup.
What is a support macro and what is an AI-drafted reply?
A support macro is a saved reply with fixed text, sometimes with merge fields like {first_name} or {order_number}. An AI-drafted reply is generated per message from the customer's email, your knowledge base, and, in some tools, live order data. A macro is retrieved. A draft is composed.
The difference matters for risk. A macro says the same thing every time, so you can audit it once. An AI draft varies, so you audit the system that produces it: the knowledge base, the rules, and the review step.
In Shopify support, macros appear under several names:
- Gmail templates
- Saved replies in help desks
- Canned responses
- Snippets in shared inboxes
AI email drafts for ecommerce sit at the other end. They read the message, pick the relevant facts, and write a response in your tone. Some tools, Heylio included, can also look up the order in Shopify before writing.
Where do canned responses win over AI?
Canned responses win when the answer is identical for every customer and wording carries risk. A saved reply for your 30-day return window takes 3 seconds to send and cannot drift. For these cases, customer support templates vs AI is not a close contest.
Three categories favor macros:
- Fixed policy text. Return windows, warranty terms, and exchange rules. One approved paragraph, zero variation.
- Legal wording. Privacy requests, chargeback notices, and regulated product disclaimers. Your lawyer approved those words. Keep them.
- Shipping-cutoff notices. 'Orders placed after 2 pm ship the next business day.' Dates and times must be exact.
Macros are also free per reply. There is no usage meter, and nothing to review before sending. Knowing when to use canned responses comes down to one test: would you be uncomfortable if the wording changed slightly between customers? If yes, use a macro.
Where do AI drafts win over saved replies?
AI drafts win when the right answer depends on the customer's order, message, or language. A saved reply cannot know that order #4812 shipped yesterday with tracking. A draft with Shopify access can, and it can answer 3 questions in one email without 3 pasted blocks.
Four situations favor AI:
- Order-specific answers. Status, address changes, and refund eligibility depend on live data.
- Multi-question emails. A customer asks about sizing, shipping to Canada, and a discount code. A macro answers one of them.
- Tone matching. An upset customer and a casual one should not get identical phrasing.
- Non-English messages. Maintaining macros in 6 languages is a chore. A draft replies in the customer's language.
The cost is review time and the need for a good knowledge base. A draft built on a stale FAQ is fast and wrong.
How do the two approaches compare?
Scored on five factors, macros lead on setup speed, error risk, and cost, while AI leads on personalization. Neither wins everything. The table below uses a small Shopify store as the reference point.
| Factor | Macros | AI drafts |
|---|---|---|
| Setup time | 1 to 2 hours for 10 to 15 replies | 1 to 3 hours to connect inbox and write the knowledge base |
| Maintenance burden | Edit each saved reply; grows with count | Edit one knowledge base; corrections feed back |
| Personalization | Low: merge fields only | High: order data, tone, language |
| Error risk | Very low if text is approved | Moderate; depends on source quality and review |
| Cost per reply | Zero beyond the tool | Metered or plan-based |
Citation capsule: For small Shopify stores, saved replies and AI drafts solve different problems. Saved replies suit the 10 to 15 recurring questions that need identical wording, such as return windows, warranty terms, and shipping cutoffs. They take about 3 seconds to send, carry near-zero error risk once approved, and cost nothing per reply. AI drafts suit emails that depend on context: a specific order, several questions at once, an unusual tone, or a language you do not write. Their trade-off is review time and dependence on an accurate knowledge base. The practical rule is simple. If the answer never changes between customers, use a macro. If the answer changes with the customer, use an AI draft built on your own policies.
What does one WISMO email look like handled each way?
A 'where is my order' email takes roughly 90 seconds with a macro and 15 seconds to review with an AI draft, because the macro forces you to look up tracking yourself. The customer writes: 'Order #4812, placed Monday, nothing yet. Where is it?'
With a macro:
- Open Shopify, find order #4812.
- Copy the tracking link.
- Paste the WISMO macro, fill in the name, order number, and link.
- Add a line about the carrier delay if one applies. Send.
With an AI draft:
- The AI reads the order number and looks up status in Shopify.
- It drafts: 'Hi Maria, order #4812 shipped Tuesday and is in transit. Tracking is here.'
- You scan it and send, or let a WISMO playbook send it.
The macro is safe but manual. The draft is faster because the lookup is the slow part. This is the one place where AI earns its keep on a routine question.
Where does Heylio fit between macros and AI drafts?
Heylio sits in the middle. Its playbooks for FAQ, WISMO, refunds, returns, and cancellations give the AI fixed rules and approved wording, so routine answers behave like macros while still pulling live Shopify data. Everything else can go to a human, and handovers are free.
Pricing is per AI resolution, meaning a conversation the AI closes on its own. A human taking over costs nothing extra. The Free plan includes 25 AI resolutions, lifetime, which is enough to test the split on real tickets before paying anything.
Heylio connects to Gmail, Outlook, Instagram DMs, WhatsApp, and a chat widget. It can look up orders, cancel, refund, apply discounts, change addresses, and issue gift cards. It also tracks your corrections and suggests knowledge base and playbook updates. It is not a replacement for your legal paragraph. Keep that as approved text.
How do you run both without duplicate maintenance?
Run both by keeping one source of truth. Write each policy once in your knowledge base, then let macros and AI drafts both reference it. Stores that keep 40 separate saved replies and a separate AI prompt end up editing the same return policy in two places.
A workable setup:
- Keep as macros: legal wording, cutoff notices, and refusal templates. Aim for 10 to 15 total.
- Let AI handle: WISMO, multi-question emails, and non-English messages.
- Review monthly: read 20 sent replies, fix the source text, and retire any macro nobody used.
- Log corrections: every edit you make to a draft is a signal that the knowledge base is wrong.
When the return window changes from 30 to 14 days, you should edit one document, then check the two macros that quote it. That is the whole maintenance routine.
Which should you pick? A short decision checklist
Pick a macro when the wording must not change. Pick an AI draft when the answer depends on the customer. Use this checklist on each question type in your inbox:
- Is the answer identical for every customer? Use a macro.
- Does a lawyer or regulator care about the wording? Use a macro.
- Does the answer need live order data? Use an AI draft or playbook.
- Does the email contain 2 or more questions? Use an AI draft.
- Is the message in a language you do not write? Use an AI draft.
- Does it happen fewer than twice a month? Skip both and answer by hand.
Next step: pull your last 25 tickets, tag each one with the list above, and test the AI-suited ones on Heylio's Free plan at https://heylio.io.