What you can automate in ecommerce customer service
An online store’s support inbox is not a hundred different questions. It is a handful of case types that repeat every day. If you want to automate customer service, start with those case types rather than the inbox, and ask the same three things of each: what data does it need, what decision, what action?
- Order status (“where is my order?”). Data: the order’s tracking. Decision: none. Action: none, just a reply with the real status. For most D2C brands this is 30–50% of all inbound tickets (the WISMO playbook).
- Address change. Data: has the order shipped? Decision: can it still be changed? Action: update the shipping address in the store.
- Cancellation. Data: order and fulfillment status, payment. Decision: can the order still be cancelled? Action: cancel and refund.
- Return. Data: delivery date, return window, items. Decision: is the case within your return policy? Action: open a return case, provide a return label, refund later.
- Warranty and defect claims. Data: purchase date, the customer’s description, history. Decision: a genuine judgement call. Action: replacement order or refund.
- Product and policy questions. Data: your product information and store policies. Action: a precise answer.
Laid out like this, it is clear where the work sits: in looking things up, deciding and executing. Typing the reply is the smallest part.
Why answer-only automation does not automate support
Many tools sold as a way to automate customer requests stop at text. A chatbot or an AI draft says “we’ll take care of your address change”, and then someone still opens the store admin, finds the order, checks whether it has shipped and edits the address by hand. The customer has a reply. You still have the case.
To automate support in an online store you need both: a reply grounded in the real order data, and the store action that closes the case. Only then does the cost per case drop, because the minutes go into looking up and clicking, not writing (the full cost math).
One example: an address change from email to store
On a Tuesday evening this email arrives:
Hi, I ordered yesterday (#10482) and have just moved. Could you ship it to 12 Hechtstrasse, 01097 Dresden instead? Thanks, Mara
Here is how Sondercase handles it:
- Classify. Sondercase recognises an address change for a specific order and matches the email to the right customer and order.
- Pull live data. Order #10482 is read from your store: paid, not yet shipped, no tracking events.
- Decide. Because the parcel has not left, the change is allowed under your rules. Sondercase prepares the new shipping address and a reply to Mara, with a confidence score and its reasoning.
- Approve and act. You release the prepared change. The address is updated in your store and Mara gets a confirmation. Several minutes in the store admin become one click.
- Escalate when it does not fit. Had the parcel already shipped, Sondercase would not guess at a change. It would hand you the case with full context.
Every case type follows the same path: message, classification, live data, decision, action. Only the risk differs per category, and that decides what you automate first.
How you stay in control of the automation
Automatic does not mean blind. With Sondercase you keep control in four places:
- Full automation per category. Replies run fully automatically wherever you switch it on. Order status can run completely on its own while warranty claims still cross your desk.
- Approval for store actions. Refunds, cancellations, address changes and returns come fully prepared, with a confidence score and the reasoning behind it. Nothing runs until you release them.
- Escalation, not guessing. When Sondercase is unsure, the case goes to a human with the order, the history and a suggested next step.
- Your rules for every action. Refunds, cancellations and address changes only run under the rules you set. Statutory consumer rights, such as the EU right of withdrawal or the statutory warranty, are treated as a floor.
The order in which to automate customer service
Start where wrong answers are rare and easy to check, and work your way towards the judgement calls:
- Order status. High volume, no discretion, every answer can be checked against tracking. The best first category.
- Product and policy questions. Pure replies with no store action. A few days in draft mode show you whether tone and content are right, then they run automatically.
- Address changes and cancellations before shipping. Clear rule, clear data: not shipped means changeable. The store action waits for your release.
- Standard returns within the window. Return case, label and reply under your return policy (more on returns).
- Warranty, defects and goodwill. Here even the replies stay in draft mode the longest. Sondercase prepares them completely; you make the call.
That is the order we followed in our own stores: order status first, warranty last (the full story).
One place for email, WhatsApp, Instagram and Facebook
Most ecommerce support still arrives by email. Sondercase connects your existing support inbox, so there is no new address to announce. WhatsApp, Instagram and Facebook are included in every plan and run on the same categories, rules and approvals. Replies go out in the customer’s language and in your brand’s tone.
Start small, then go automatic
Automating customer service pays off when the case is actually done afterwards: a reply with real data, the action in your store, traceable and under your rules. Start in draft mode, switch categories to automatic one by one, and keep the judgement calls for yourself. Store actions wait for your approval.
To see what this looks like with your own store’s requests, try the live demo: enter your store domain, Sondercase analyses your store and resolves a real support case, no signup needed. After that you can try it free for 14 days, no credit card required. How the AI behind it works is explained on AI customer service.