Industries/ Retail/ Order and returns assistant
Where is my order, will it fit, can I return it: answered from real data
We build an assistant for your website, WhatsApp and Instagram that looks up the customer's actual order, applies your actual returns policy and reads the size guide for the product in front of them. Anything unusual goes to a person on your team with the whole conversation attached.

- Used by
- Your customers, customer service and the e-commerce manager
- Takes the place of
- An inbox full of order-status emails and unanswered DMs
- Connects to
- Your store platform, couriers, helpdesk and messaging apps
- Built with
- Our ai receptionists & chat service
Where it hurts today
A normal week, before it exists
Open the customer service inbox on the Monday after a sale weekend and most of it is the same handful of questions. Where is my order. Has my refund gone through. Can I return this if I have worn it once. Does the dress come up small. Each one takes a few minutes: find the order, open the courier's site, check the policy, write a polite reply. Multiply that by the weekend and Monday has gone.
The questions arrive when nobody is there to answer them. The evening after a sale email, the Sunday before a birthday, late at night when someone is deciding between two sizes. On Instagram, a message asking whether the medium is back in stock sits unread until Tuesday, by which time the customer has bought something similar elsewhere. Nobody chose to ignore it. There was simply nobody on shift.
You may have tried a chatbot already. The one you tried answered from a list of canned replies, could not see a single real order, and sent anyone with a genuine problem round in circles. Customers learned to type agent until something gave way. Your team ended up answering the same questions anyway, plus the complaints about the bot, and it was quietly switched off.
What it does
Order and returns assistant, feature by feature
What an e-commerce chatbot does for a retailer when it is wired into the systems you already run, rather than written as canned replies.
- 01
Real order lookups and tracking
The customer gives an order number and the email or postcode it was placed with. The assistant reads the order from your store and the tracking from the courier, then says where the parcel actually is and what happens next, rather than a generic delivery estimate.
- 02
Your returns policy, applied to their order
It knows your returns window, which items are final sale, how exchanges work and whether returns are free. Asked about a specific order, it checks the delivery date and the items, says what that customer can do, and starts the return where your policy allows it.
- 03
The size guide for the product in question
Asked whether a dress comes up small, it reads that product's measurements and the fit notes your team has written, and answers for that item. It does not guess from a generic chart, and when the notes say nothing about fit, it says so plainly.
- 04
Live stock and back-in-stock alerts
On Instagram or WhatsApp, a customer asks whether the medium is back. The assistant checks live stock, sends the product link if it is, and offers a back-in-stock alert if it is not, with the customer's permission to contact them recorded.
- 05
A clean handover to your team
Damaged parcels, refund disputes, address changes after dispatch, anything angry and anything outside policy go to a person. The conversation, the order and what the customer has already been told land in your helpdesk, so nobody asks them to start again.
- 06
A list of what it could not answer
Every question it handed over or could not answer is grouped by theme each week. When several people ask about the same missing care detail or confusing delivery option, your team sees it and fixes the product page, not just the reply.
A day with it
How the day runs instead
A normal evening and the morning after, with the assistant answering alongside your customer service team.
- 6.40pmCustomer on the bus
Asks on WhatsApp where order 4821 is. The assistant asks for the postcode, checks the order and the courier's record, and replies that it is out for delivery tomorrow, with the courier's tracking link and the delivery window the courier has given.
- 9.15pmShopper choosing a size
Asks the website chat whether the linen dress runs small. The assistant quotes the bust and waist measurements for sizes 12 and 14 and your team's note that the cut is generous, then links straight back to the product page.
- 11pmCustomer with a broken mug
Sends a photo through Instagram. The assistant apologises, records the order and the photo, says a person will reply in the morning when the team is back, and opens a ticket in the helpdesk marked damaged in transit.
- 8.30amCustomer service lead
Opens the helpdesk to find the overnight handovers waiting, each with the order, the photos and the full conversation. A replacement mug is sent before nine. The routine questions from the evening were already answered and need nothing from the team.
- FridayE-commerce manager
Reviews the week's unanswered questions. Several people asked whether a coat is waterproof, and the product page never said. The fabric detail goes on the page that afternoon, and from then on the assistant can answer it too.
What it connects to
Works with what you already run
An assistant is only as useful as what it can see. These are the four connections that let it answer from real data.
Online store platform
Orders, line items, delivery dates, refunds and product details, including size guides and fit notes, read through your store's interface. Where your policy allows, it can also start a return or add a delivery note before dispatch.
Couriers and shipping software
Tracking events from your carriers or shipping platform, so the answer to where is my order comes from the courier's own record rather than an estimate printed at checkout.
Helpdesk or shared inbox
Handovers arrive as tickets in the helpdesk your team already uses, for example Gorgias or Zendesk where your plan includes API access, with the conversation, the order and any photos attached.
Messaging channels
Website chat, the WhatsApp Business Platform and Instagram messages all reach the same assistant, with one set of answers and one set of handover rules behind them, rather than three separate tools.
The first version
Start small, then grow it
A first version should be useful on day one and small enough to get right. This is where we would usually draw the line, and what tends to follow once people are using it.
In the first version
- Order status and tracking answers in your website chat, from live store and courier data
- Your returns policy applied to the customer's actual order
- Size guide and fit note answers for every product that has them
- Handover to your helpdesk with the full conversation and order attached
- A weekly list of questions it could not answer, grouped by theme
Once it is in use
- WhatsApp and Instagram messages handled by the same assistant
- Returns and exchanges started in the chat, with labels sent automatically
- Answers in the customer's own language from your English policies
- Back-in-stock alerts offered in the chat, with permission recorded
Buyer's guide
Before you commission one
Four decisions worth making before you commission an e-commerce chatbot, including whether to commission one at all.
- 01
Your helpdesk's built-in AI or a custom assistant
Many e-commerce helpdesks now sell an AI agent as an add-on. If you run one store platform, one courier and a straightforward returns policy, that add-on is often the better buy. A custom assistant earns its place when answers depend on several systems, such as tills, marketplaces and more than one courier, or when you want to own the rules, the prompts and the conversation data.
- 02
What to gather first
A few weeks of real customer emails and messages, with personal details removed; your returns policy exactly as customers see it; size guides and fit notes per product; your courier accounts; and a short list of things the assistant must never say. Where two of your documents contradict each other, we will find it, and someone on your side will need to decide which is right.
- 03
Who owns it inside the business
The customer service lead should own the answers, the handover rules and the weekly review of what it got wrong. The e-commerce manager owns the product data it reads, because a missing fit note is a product page problem. Without a named owner, the gaps list grows and nobody acts on it.
- 04
How you will know it is working
Watch how many conversations end without a handover, how often the same customer comes back about the same order, and how quickly your team picks up the handovers it does make. Read a sample of conversations every week. Complaints about the assistant itself are the clearest warning, and a shrinking unanswered list is the clearest sign of progress.
Terms worth knowing
The words you will meet
- WISMO
- Short for where is my order, the question that fills most retail support inboxes. It is the obvious first job for an assistant because the answer exists in your systems already. Answering it well needs live order and courier data, not a link to a generic tracking page.
- Retrieval-augmented generation
- Often shortened to RAG. Before answering, the assistant looks up the relevant order, policy or product page and answers from that, rather than from what the AI model absorbed in training. It is the main defence against an assistant inventing a returns policy you do not have.
- WhatsApp Business Platform
- The official route for businesses to message customers on WhatsApp through an approved provider. Since 15 January 2026, Meta's terms bar general-purpose AI assistants from the platform, while assistants that serve a business's own customers, such as order and returns help, remain permitted. We check the current terms before building.
Handled with care
The details that matter here
An assistant speaks for your shop to customers you have never met, often late at night. These are the details we handle with particular care.
Checking who it is talking to
Before it reveals anything about an order, the assistant asks for the order number together with the email or postcode it was placed with. It never reads out a full address or payment details, and we agree with you what it may confirm to someone who cannot verify.
What it must never promise
We write down with you what the assistant may not do: offer discounts, approve refunds outside your policy, promise delivery dates the courier has not given, or argue about a complaint. Those rules are tested against your real past questions before it goes live, and again after every change.
Labelled as an assistant, with a person reachable
It says it is an AI assistant from the first message. In the EU, the AI Act has required that disclosure for chatbots since 2 August 2026, and we apply it in every market. Customers can ask for a person at any point, and it says honestly when your team is next working.
Conversation records and retention
Chats contain names, addresses and sometimes photos, all personal data under UK GDPR and EU GDPR. We agree with you how long conversations are kept, who on your team can read them and where the data is processed, and we set it up so a customer's request to delete them can be honoured.
Guides
Worth reading first
Questions
What buyers ask first
Can an AI chatbot look up real orders and tracking information?
Yes, if it is connected to your store platform and your couriers, and that connection is most of the work. Once the customer has verified the order with an email or postcode, the assistant reads the order, the items and the courier's tracking events, and answers from them. Where a courier does not offer tracking data, it says what it knows and passes the conversation on rather than guessing.
Will an e-commerce chatbot give away refunds or discount codes?
Not unless you decide it may. The assistant can only take the actions we connect it to, each with limits you set. A sensible starting point is letting it start a return within policy and nothing more. Refunds outside the window, goodwill gestures and discount codes stay with your team, and the assistant hands those conversations over rather than improvising an answer.
Does a chatbot have to tell customers it is AI?
In the EU, yes. Article 50 of the AI Act has required, since 2 August 2026, that people are told when they are interacting with an AI system unless it is obvious. Rules elsewhere differ, but we label the assistant clearly in every market anyway. Customers trust it more when they know what it is and can see how to reach a person.
What if the chatbot gets an answer wrong?
It can, and we plan for that rather than pretend otherwise. It only answers from your store data, courier records and the policies you give it, and it is set to hand over rather than guess when those do not cover the question. Before launch we test it against a set of your real past questions. You can read every conversation, and each mistake becomes a test for the next version.
Will an AI chatbot replace our customer service team?
No. It takes the repetitive questions, mostly order status, sizing and returns, so your team can spend its time on damaged parcels, disputes and customers who need judgement. The aim is for the team's work to change: fewer copied-and-pasted replies, more time on the cases that need a person, and handovers that arrive with the details already gathered.
Send us the questions your inbox answers every day
Forward us a week of customer service emails, with personal details removed. We will tell you which ones an assistant could answer from your data and which should stay with your team, and we reply within 48 hours.
More for retail
What else we build for retail and e-commerce
Online stores that convert
Fast product pages, search that copes with typos, delivery costs shown early and a checkout with nothing extra in it. Built on Shopify with a custom storefront, or on an open engine when your pricing and rules do not fit.
One stock count everywhere
The tills, the website, the marketplaces and the warehouse syncing to one agreed stock figure, with clear rules for which system wins when two disagree, so overselling stops being a weekly apology to customers.
Shopping and loyalty apps
An app for your regulars with saved baskets, loyalty points that work in the shop and online, back-in-stock alerts, and notifications people have actually asked for rather than the kind they mute after a week.
Store and channel dashboards
Yesterday's takings, margin, stock and returns by shop and by channel, ready each morning on a phone or an office screen, so a bad week gets spotted while there is still time to do something about it.
Returns and back-office tools
A returns desk that records a reason every time, refunds handled in one place, click-and-collect pick lists for the shop floor, and the small internal screens that replace the spreadsheet everyone is afraid to sort.
Everything we build for retail
The same idea, in other sectors


