Three “where is my order?” emails landed overnight, and yesterday’s orders are not even packed. A cart worth a couple of hundred dollars has sat abandoned since Monday, and you have been meaning to ask last month’s customers for a review since, well, last month.
None of that is fixed by working harder. It is admin, and admin is the first thing to fall over when the orders are actually coming in. Here are five ecommerce workflows you can hand to an AI agent — what each does, what stays with a human, and the realistic first step.
What an AI agent actually does
Most people have met AI as a chat window: you ask, it answers. An agent is the next step — connected to your store platform, email, SMS and inventory, and allowed to take an action, not just reply. It sends the cart reminder, answers the tracking question, attaches the return label. It does the legwork and drafting; you keep the decisions. For more, see what AI agents are.
Ordinary automation follows a fixed script and falls over when reality gets messy. An agent handles the messy middle — reading “is my order coming today?”, checking the live tracking, and answering in your tone. That is agentic AI, and why you keep a human checking the output while you learn the odd ones.
Ecommerce automation starts with the one job nobody wants to do
The stores that get something out of this start with one task: repetitive, mostly text, and low-risk if it needs a second look. Three worries come up in nearly every conversation, and all three are fair.
- “It will sound like a robot.” It will if you let it. Feed it your real past replies, keep messages short, and read the first weeks before anything sends by itself.
- “It will become a big expensive project.” Only if you scope it like one. A single workflow plugged into tools you already pay for either earns its keep in the first fortnight or you switch it off.
- “I will lose the personal touch.” That is a design decision: the agent takes the chase and the reminder, and hands over the moment a conversation needs judgement.
We have covered where to start with AI automation in more detail, and if you run a local trade or service business, we wrote the same guide for tradies.
1. Abandoned carts: win back the sale you already earned
An abandoned cart is a customer who already said yes, then got distracted. When a cart goes quiet, the agent sends a short nudge — within an hour or two, then once more a day later — mentioning the items and making it one tap to finish. Drafts come to you for approval, and the sequence stops the moment the customer replies or buys.
What to keep human: discounts — an agent that quietly drops the price to force a sale teaches customers to wait for a code. First step: one cart email, one timing, drafts only.
2. Order status: answer “where is my order?” without living in your inbox
“Where is my order?” is the most common support question in ecommerce, and it is nearly always answerable without you. The agent pulls the live tracking, translates it into plain English, and can send proactive updates — dispatched, out for delivery, delivered — so the question never gets asked. What lands back with you is the small queue that genuinely needs you.
What to keep human: anything a tracking link cannot answer — lost parcels, damage, refunds, an angry customer. First step: turn on one proactive “your order has shipped” message with the tracking link.
3. Review requests: ask every customer, not just the happy ones
Reviews do a lot of the work in whether a new customer trusts you, and most of the ones you never got were simply never asked for. When an order is marked delivered, the agent sends a short request a few days later with a direct link, and can draft replies to reviews as they come in.
One compliance point: ask every customer, and never offer a discount, freebie or prize in exchange for a review. Google’s Business Profile policies prohibit incentivised reviews, and a profile in breach can lose the ability to collect new reviews for a period. What to keep human: your reply to the tough reviews. First step: automate the request for one order type, with a direct link.
4. Back-in-stock and inventory alerts: turn the waitlist into sales
Every “notify me when it’s back” is a customer who has already told you they want to buy. The agent tracks the items people asked about and, the moment one is back, sends a short personal note — not a marketing blast — with a direct link. Cap alerts against actual stock so nobody is told “it’s back” and finds it gone.
What to keep human: which items deserve a personal alert and which should just quietly restock. First step: pick one fast-selling product, connect the waitlist, and send alerts for that one item.
5. Returns and support triage: sort the messy middle
Returns are where a good store is won or lost, and every request is a little different. The agent reads the incoming message and routes it: a simple “how do I return this?” gets the label and process in one reply, while a faulty-item complaint gets gathered — order number, what is wrong, photos — and handed to you ready for a decision.
What to keep human: the decisions — whether to refund, replace or credit, and any goodwill gesture. First step: write down your three most common support questions and script the answer for one.
How the five fit together (a Melbourne example)
Picture a homewares store run from a warehouse in Footscray. A $180 cart goes quiet Tuesday night; the agent nudges, the customer asks “does the vase ship to Darwin?”, and the order completes. Thursday, a “where is my order?” email is answered from live tracking with no human involved. Twelve people are waiting on a sold-out linen throw — when it restocks Friday, the agent alerts the first eight and holds the rest. That is one thread, not five products, and stitching it together is most of what ecommerce automation involves.
What to watch out for before you switch anything on
- Keep a human in the loop for the first fortnight — everything drafts-for-approval.
- Write down what the agent must never do — refunds, discounts, timeframes, payment.
- Know where your customer data goes — which tool holds it and who can see it.
- Leave an obvious way to reach a person within a message or two.
- Follow the Australian messaging rules. The Spam Act 2003 applies regardless of who pressed send — you need consent, a working unsubscribe, and action within five business days. Keep transactional updates separate from marketing.
- Make it sound like you — read the drafts out loud.
Where to start this week
- Keep a list for one week. Every repetitive admin task you do, write it down. The list is the audit.
- Pick the single most annoying one. Not the biggest — the one that makes you sigh.
- Run it as drafts for a fortnight. Let it write, you press send. When you stop editing, let it send on its own.
Most online stores do not need a platform — they need one annoying job handled properly. If you would rather not work out which task to start with, Web Intelligenz builds AI agents and workflow automation for Melbourne ecommerce and local businesses, and you are welcome to book a short call to talk through which workflow is worth doing first.
Frequently asked questions
What is ecommerce automation?
Handing a task that used to need a person reading, writing or deciding to a system — sending the cart reminder, answering the tracking question, sorting a support message.
How does agentic AI differ from traditional automation?
Traditional automation follows fixed rules. Agentic AI is given a goal and works out the steps, so it copes with a vague question or an unexpected reply — which is why it needs explicit boundaries about what it must never do.
Sources: Australian Communications and Media Authority, “Avoid sending spam” and the Spam Act 2003 consent and unsubscribe rules; Google Business Profile Help, “Prohibited & restricted content” — fake engagement and incentivised reviews (both accessed September 2026). No statistics or survey figures are quoted in this article.
