Ecommerce: automate post-sale support, keep it human

In an online shop, the sale doesn't end when the customer pays. It ends when the parcel arrives in one piece, the size fits, and nobody's left hanging with a half-answered question. That part — the after-sale — is the least glamorous and the most time-hungry: "where's my order?", "how do I return this?", "will the 40 fit me?". They're messages that repeat every day, nearly always the same, landing at all hours. If you run a small ecommerce and you feel like you spend more time on orders you've already sold than on new customers, this is about winning those hours back without your shop losing the personal touch that sets it apart.

The "where's my order?" that never ends

If you had to bet, most of the messages you get after a purchase are versions of the same thing: where's the parcel and when does it arrive. The trade calls it WISMO (where is my order), and it has one feature that makes it perfect to automate: the answer already exists. Nothing needs deciding — it needs looking up. The tracking number is in your platform, the live status sits on the courier's site, and joining those two things is exactly what a customer can't do alone and what interrupts you twenty times a day.

An AI agent wired into your shop and the courier's tracking bridges that gap without you. The customer asks on WhatsApp at eleven at night, the agent finds their order by email or number, checks the status and replies with real information: "Your order left yesterday, it's in Seville, and you'll have it tomorrow before 2 pm." Not a "let me check" that dumps the job back on you, but the full answer. And if the parcel is stuck or hasn't moved in three days, that it does pass to you — which is precisely when a human is needed.

Returns and exchanges without losing your afternoon

The return is the moment the customer is most anxious and you're most tied to the process. For them it's a hassle ("do I pay for postage? when do I get my money back?"); for you it's a chain of almost identical steps: check it's within the window, generate the label, log the return and flag the refund timeline. None of it calls for judgement — it calls for following your rules to the letter.

That's where an agent earns its keep. You give it your policy — "30 days, unused, refund in 5–7 working days" — and it applies it: asks only what's needed, generates the return label, logs the case and tells the customer exactly what happens and when. A size swap, which is a return and a fresh sale rolled into one, it handles just as smoothly: it collects what didn't fit and sets up the right size going out. You keep what falls outside the script: the item that arrived broken, the out-of-window return you want to be generous with, the complaint that smells like trouble. Those don't get automated, and rightly so.

Sizing and product questions: the sale you haven't closed yet

There's a kind of after-sale message that's really pre-sale in disguise: "does this run large?", "is the blue like the photo?", "is it dishwasher-safe?". Answering them fast and well pays off twice: you head off a return before it happens and, often, you close a purchase that was hanging in the balance. Every sizing doubt cleared in time is a parcel that doesn't come back.

An agent with access to your product page and size guide can answer these instantly, and with nuance a drop-down can't offer: "This one runs a touch small; if you're between a 40 and a 42, go for the 42." One caveat, though: the agent should speak to what it knows — measurements, materials, what the listing says — and not promise an "it'll fit you perfectly" it can't possibly know. Calibrated well, it helps people choose right; calibrated badly, it creates the very returns you were trying to avoid.

What NOT to automate (yet)

An after-sale agent isn't a magic button, and saying so plainly saves us both a bad experience. There are three situations where it's worth pumping the brakes:

• When money is involved beyond your rules. An extra refund, a goodwill discount, reshipping a lost order for free — those are your calls. The agent can tee up the case, but the final word is yours.
• When the customer is genuinely angry. A serious complaint handled badly by a bot does more damage than the original problem. The healthy rule is that the agent reads the tone and hands it to you before it puts its foot in it.
• When your data is a mess. If stock doesn't match, listings are half-finished or order status doesn't update, the agent will just repeat your mistakes faster. Tidy the house first — even with a decent spreadsheet — and automate after.

What this looks like set up in your shop

You don't need to switch platforms or build anything exotic. The agent connects to what you already use — Shopify, WooCommerce, PrestaShop — and to your messaging channel, and works from three things: your order status, your returns policy written as rules, and your size guide. For the first few weeks you run it in "propose and I approve" mode: it drafts the reply, you sign it off, and you learn to trust it. Once you see it nail the usual cases — tracking, standard returns, sizing — you loosen the reins and keep only the exceptions for yourself. You approve, they execute.

The result isn't a colder shop — quite the opposite. Customers get an instant answer to the same old questions, and you show up — with time and a clear head — exactly when something goes wrong and your judgement matters.

If your after-sale fills up with tracking chases, returns and sizing doubts, that's exactly what our receptionist handles: it answers on WhatsApp, web and email, checks the order status and applies your rules, and passes to you whatever needs your decision.