Measured across ReplAi's merchant cohort in Q1 2026, 80% of messages were fully automated with no human involvement. For stores doing 1,000+ orders per month, the sustained range is 70-80%. The honest answer, then: most tickets, but not all, and the remainder genuinely needs people.
Where does the 80% figure come from?
Vendors throw automation percentages around loosely, so here is exactly what ours means. Across the ReplAi merchant cohort in Q1 2026, 80% of inbound messages were resolved end to end by the AI: understood, answered, and closed without a human ever touching the conversation. The cohort spans merchants like Bronze, Blaze Sportswear, Ivory, Gioia, Basic Look, and Glo Swim, collectively generating 4,000+ DMs per day. To date, ReplAi has processed more than 500,000 customer messages.
Two caveats we would want to see from any vendor, so we apply them to ourselves:
-
It is a cohort measurement, not a guarantee. Your mix of ticket types, policies, and catalogue complexity moves the number. Stores at 1,000+ orders per month reliably land in the 70-80% band.
-
"Automated" means fully resolved, not "the bot replied once before a human cleaned up." Conversations a human takes over count on the human side of the ledger.
Why is so much of support volume automatable?
Because support volume is brutally concentrated. A handful of intents dominate:
-
WISMO (where is my order): commonly 25-50% of ecommerce ticket volume per industry reporting, and entirely answerable from live Shopify order data; see our WISMO automation guide
-
Product questions: stock, sizing, materials, price, answerable from the catalogue
-
Policy questions: shipping times, return windows, payment options
-
Returns and exchanges: rule-based workflows the AI can execute in Shopify, as covered in can AI process a return on Shopify
-
Order placement: including COD orders taken in conversation, plus abandoned cart follow-ups (relevant given roughly 70% average cart abandonment, per Baymard Institute)
Each of these is high-frequency, low-judgment, and data-backed. That is the automation sweet spot, and together they cover the large majority of what lands in an ecommerce inbox.
What is in the 20-30% that still needs humans?
Being specific about the remainder matters more than celebrating the majority.
| Category | Example | Why AI should hand off |
|---|---|---|
| Emotional escalations | Furious customer, third delayed order | Empathy and goodwill are judgment calls |
| Policy exceptions | Return request one week outside the window | Exceptions are brand decisions |
| Delivery failures | "Marked delivered, never arrived" | Investigation and compensation judgment |
| Unusual requests | Bulk/B2B orders, collabs, press | Off-script by definition |
| Ambiguity the AI flags | Contradictory order details | Better to escalate than guess |
A trustworthy system does two things with these: escalates early rather than looping, and hands the human full context (order, history, what the customer wants) so resolution starts warm. In ReplAi, that handoff happens in the same inbox, with the AI toggled off for that conversation.
Does a higher automation percentage always mean better?
No, and this is worth internalising. A tool can inflate its automation rate by refusing to escalate, trapping customers in deflection loops. That gains you a metric and costs you customers. The right target is the highest automation rate at which resolution quality holds, which in our cohort experience settles around that 70-80% band for real stores with real edge cases.
The better questions to ask of any tool (ours included):
-
What counts as "resolved" in your metric?
-
How fast does escalation happen when the AI is unsure?
-
Can my team take over mid-conversation without the customer repeating themselves?
-
What does the automation rate do to response time on the human-handled remainder?
That last one is underrated: when AI absorbs the routine 70-80%, your humans answer the hard 20-30% dramatically faster. Stores doing 1,000+ orders per month on ReplAi free up 4-8 hours of support team time per day, and much of that time flows back into the escalations.
What does 80% automation change in practice?
Three operational shifts:
-
Response time collapses. ReplAi answers in under 14 seconds on average, 24/7. The customer messaging at midnight and the one messaging at noon get the same experience.
-
The team's job changes. From copy-pasting tracking numbers to handling exceptions and improving the system. Fewer, better human conversations.
-
DMs become a sales channel. Fast answers convert. Against a baseline DM close rate of about 3%, ReplAi's pilot cohort saw DM sales lift of +75% (bottom quartile), +150% (median), and +300% (top quartile), with fashion brand Bronze at 13x in a single month as the top outlier.
How can you test the number on your own store?
Do not take a cohort statistic on faith; generate your own. ReplAi's free plan (100 replies per month) exists for exactly this: connect your store (one-click Shopify install, about 10 minutes, real-person onboarding included), run it on live conversations for two weeks, and read your own automation rate off the dashboard. Then size a plan against reality on the pricing page, or book a demo first if you want a walkthrough with your ticket mix.
Frequently asked questions
Is 80% automation realistic for a brand-new store? The intent mix is what matters, not store age. New stores actually skew toward highly automatable pre-sale and shipping questions. What new stores lack is conversation volume, so percentages are noisy week to week. Judge the trend over a month, not a Tuesday.
Why do you quote 70-80% for high-volume stores but 80% for the cohort? Different measurements. The 80% figure is the Q1 2026 average across the whole merchant cohort. The 70-80% band is what we consistently observe as the operating range for stores doing 1,000+ orders per month, quoted as a range because ticket mix varies by store. We would rather give you both numbers with definitions than one impressive number without.
Can I choose which ticket types the AI handles? Yes, and you should. Sensible rollouts start with WISMO and product questions, add returns once trust is built, and keep categories like refund disputes permanently human. ReplAi's per-conversation AI toggle and escalation rules make the boundary explicit rather than emergent.
What happens to my automation rate during a crisis, like a courier meltdown? Volume spikes and the mix shifts toward exactly the tickets AI handles best: "where is my order." Automation absorbs the flood while your team works the genuine failures. This is when flat per-reply pricing also matters, since per-ticket billing models turn the same crisis into an invoice; see our note on that in Gorgias pricing explained.