E-commerce automation that runs your store's busywork, not slide decks about it

Where the hours actually go in a DTC store

Talk to any ops lead at a scaling store and the same list comes up. Customers asking where their order is. Return and refund requests piling up in the inbox. Inventory that says one thing on Shopify and another on Amazon and a third in the warehouse. Review requests that never go out because nobody has time. Abandoned carts that just quietly leak revenue. None of it is hard. All of it is constant, and it scales linearly with orders, so the busier you get, the more people you have to throw at work a machine should be doing.

The trap is hiring your way out of it. Another support rep, another VA, another spreadsheet owner. That buys you a few months before the volume catches up again, and now you have a bigger payroll doing the same repetitive work. The better move is to automate the jobs that are high-frequency and rule-based, and keep your people for the parts that need a human: the angry customer, the wholesale negotiation, the product decision.

We don't automate everything at once. We start with the one workflow that has the highest volume and the clearest rules, prove it in production, then move to the next. That is usually WISMO or returns, because in most stores those two categories alone are half the support queue.

The first five workflows we automate for a store

These are the jobs that pay back fastest, in roughly the order we tackle them. Each one is a real pipeline wired into the tools you already run, not a chatbot bolted on the side. We build them in n8n when there's branching and real data to move, in Make or Zapier for simpler app-to-app hops, and in custom code when an API is stubborn or a store is high-volume enough to need it.

The point of starting here is that every item on this list is countable before we build. We can look at your ticket volume, your return rate, your cart-abandonment number, and tell you what an automation is worth in hours or recovered revenue. If the math doesn't clear the bar, we say so.

• WISMO (where-is-my-order): an assistant reads the question, pulls live tracking from your carrier and Shopify, and replies with the real status in seconds, day or night.
• Returns and refunds: policy checks run automatically (in the window, eligible item, under your auto-approve threshold), the label is generated, and the refund is issued through Stripe or Shopify Payments once conditions are met.
• Inventory and order sync: stock levels stay consistent across Shopify, Amazon, and your 3PL, with low-stock alerts and auto-reorder triggers so you stop overselling.
• Review and UGC requests: post-delivery, the right customer gets the right review or photo request at the right moment, wired to Judge.me, Okendo, or Loox.
• Abandoned cart and win-back: dynamic flows that recover carts and re-engage lapsed buyers via Klaviyo, personalized on what they actually looked at, not a generic blast.

How a WISMO or returns automation actually works

Take the most common one, order status. Today a human opens the ticket, copies the order number, checks Shopify, opens the carrier's tracking page, reads the status, writes a reply, and closes the ticket. Eight minutes if nothing's weird. We rebuild that as a pipeline: an LLM classifies the incoming message as a status request and extracts the order reference, the automation looks up the order in Shopify and pulls live tracking from the carrier API, and it drafts a reply grounded in the real data. On the clear cases it sends and closes on its own. On the edge cases (a package marked delivered but the customer says it never arrived) it escalates to a human with all the context already gathered.

Returns work the same way, but with money involved, so the guardrails are tighter. The automation checks the request against your written policy: is the order inside the return window, is the item eligible, is the refund amount under the threshold you set. Only when every check passes does it generate the label and issue the refund. Anything over the threshold, or anything unusual, stops for a human to approve. Every action is logged and idempotent, so a retried request never refunds twice, and you can reconstruct exactly what happened on any order.

This is the part that separates a real automation from a toy. It reads messy inputs, applies your rules, writes back to your systems, and knows when to stop and ask. The customer sees a fast, correct answer. You see a queue that used to need three people now needing one, watching a dashboard.

The stack, and why nothing gets ripped out

We plug into the tools you already run. Your store on Shopify or WooCommerce, your support in Gorgias, Zendesk, or Re:amaze, your email and SMS in Klaviyo, your reviews in Judge.me or Okendo, your fulfillment in ShipBob, ShipStation, or a custom 3PL. The automation becomes the connective tissue between them, doing the copy-paste and lookups a human does by hand today. You keep your stack; we make it talk to itself.

For the orchestration layer we lean on n8n, which we can self-host so your customer and order data stays in the EU, a real consideration if you sell into Germany and the rest of Europe. Simpler triggers run on Make or Zapier where that's the cheaper, faster option. When a store's volume or a carrier's API demands it, we write custom code. The tool follows the job, never the other way around, and you own what we build: the workflows, the logic, the credentials, all of it.

• Storefronts: Shopify, Shopify Plus, WooCommerce, and headless setups via API.
• Support: Gorgias, Zendesk, Re:amaze, Intercom, with AI drafting and auto-resolution.
• Email and SMS: Klaviyo, Postscript, and Attentive for flows and win-back.
• Reviews and UGC: Judge.me, Okendo, Loox, triggered on delivery events.
• Fulfillment and inventory: ShipBob, ShipStation, Amazon, and custom 3PL feeds.
• Payments: Stripe and Shopify Payments for automated, policy-gated refunds.

What it's worth, and when we'll tell you to wait

The math on a busy store is usually blunt. A support inbox handling 400 WISMO tickets a month at eight minutes each is 53 hours, more than a full-time week and a half, spent on a question a machine answers in seconds. Automate returns on top of that and you often free a second person entirely. On the revenue side, a well-built Klaviyo cart-recovery flow routinely recovers a meaningful share of abandoned carts, and that number lands straight on the top line. We scope every build to the number first: how many runs, how many minutes each, what a mistake costs. Most store automations reach production in two to four weeks and pay back inside the first one to three months.

It isn't always the right move, and we'll say so before you spend. If you're doing 30 orders a month, a full returns automation costs more to build and maintain than the hours it saves; a good policy page and a template do the job. If your catalog and rules change every week, the automation needs constant rework and the maintenance eats the gain. And some things stay human on purpose: a high-value wholesale inquiry, a genuinely upset customer, a fraud edge case. We automate the repetitive 80% and hand your team the tricky 20% with better context than they had before. Because we tie our fee to the automation actually running in production, we only take on the workflows where the ROI is real, which is exactly why we're honest about the ones that aren't worth it yet.