--- title: "E-commerce Automation: 7 Order Stages | JustAutomate" description: "We connect the shop, warehouse, courier and accounting. One order from click to return, stock-outs and complaints included. Workshop EUR 3,500 a week." url: "https://justautomate.ai/en/services/ecommerce-automation/" lang: en --- Process automation # E-commerce automation from click to return One order passes through the platform, the warehouse, the courier, accounting and the customer's inbox. We connect those systems so the data moves between them on its own, and a person steps in where a decision is needed: a stock-out, a partial shipment, a complaint. We start with one stage, the one that touches every single order. Book a free consultation [See how the workshop works](https://justautomate.ai/en/services/process-workshop/) - 7 stagesthe distance one order travels from click to complaint, and each stage can be taken over on its own - 5 systemswhat usually stands along that road: the platform, the warehouse, the courier, accounting and customer messaging - 2 to 4 weeksfrom the workshop to the first stage running in production, switched on next to your current process The problem ## How many times does one order pass through human hands? An order lands on Sunday at 23:15\. On Monday someone logs into the panel, pulls the list, checks stock in a spreadsheet, looks whether the invoice exists yet, copies the address into the courier system, pastes the tracking number back into the shop and writes to the customer. Seven hand-offs of the same data on one order, and not one of them needs a decision, only attention. At thirty orders a day that is more than two hundred such actions every day. They grow in a straight line with the number of parcels, so peak season gets solved by hiring rather than by process. The worse part comes around the two hundredth hand-off, when a tracking number goes onto the wrong order and you hear about it from the customer. Exceptions are a separate matter, because their clocks start outside your company. A customer files a return and the refund deadline starts running that day, set by the consumer law of the market you sell into, whether or not anyone has opened the parcel yet. A parcel lost at the courier and a complaint work the same way. The clock starts at the moment of the filing, not at the moment someone spots a box on the warehouse floor. - Tracking numbers pasted by hand, so the customer learns about the shipment a day late - Stock on the platform and stock in the warehouse drift apart until the next export - A missing item is found by the packer at the bench, not by the system at checkout - A return sits in its box while the refund deadline runs from the day it was filed Why now ## Why can this go in one stage at a time? Every stage of the order path already has its own point of contact that can be connected on its own: the new order event, a shipping status change, a stock movement, a return filed through a form. You do not have to take the whole shop at once. The first stage goes into production while everything else keeps working exactly as it does today. The courier layer has already been assembled for you. Order systems such as BaseLinker hold the integrations with InPost, DPD, DHL and GLS, so we connect to one place instead of four carrier panels one after another. Marketplaces work the same way: Allegro, Amazon and eBay come in through the same point as your own shop. The third change is about questions, not about flows. A rule will move an order from one stage to the next, but it will not tell your support team why one particular parcel has been sitting at the courier for three days. BOND, our company AI agent, reads the shop, the warehouse and the shipping system at once and answers from company data in 5 to 10 seconds. What changes ## What exactly stops needing a person? - ### An order reaches packing with nothing retyped A paid order reserves stock, creates the picking document and the confirmation for the customer. The warehouse gets a finished list instead of an export somebody has to check by hand first. One event on the platform starts three documents in three systems - ### The tracking number reaches the customer without copy and paste The label is built from the order data, and the tracking number goes back to the platform and into the message to the customer. A status change at the courier updates the order, so support stops looking for it in the carrier panel. InPost, DPD, DHL and GLS through one connection point, not four panels - ### One stock figure instead of four versions of it The shop, the warehouse and the marketplaces read the same stock, updated when goods actually move. When a line hits zero the offer comes down before somebody buys what you do not have. Updated on the movement of goods, not by an overnight export to a file - ### Every exception gets a person and a deadline A stock-out, a partial shipment, a parcel lost at the courier and a complaint each have a written path and land on a named person with a date. The automation runs the standard path and recognises by itself when an order is not on it. Four exceptions written down in the workshop, before the first line of code - ### A return keeps its deadline from the first hour A return filed through the form creates the label, sets the status and opens a task with a deadline counted from the day of filing. The refund starts once the goods are booked in or once there is proof of posting, whichever your policy allows. The deadline counted from the filing date, on a task with a named owner - ### Order status without opening three panels The question what is happening with order 12345 gets asked in a plain sentence, and the AI agent assembles the answer from the shop, the warehouse and the shipping system. Nobody logs into three systems in turn. An answer from company data in 5 to 10 seconds Proof ## Do you run a shop like this yourselves? Yes, our own. The clothing brand our founders built sells from a PrestaShop store wired into an order system and an ERP, and sixteen people there use an AI agent, the warehouse included. Below is the only number we have from measurement, together with what it does NOT prove: it talks about return on ad spend, not about the time it takes to handle an order. Nobody here measured order handling time before the rollout, so there is nothing to compare it against. Built for ourselves. These are not client deployments but implementations inside the founders' own companies. We say so plainly, because proof you cannot verify is not proof. - ![Surf Inc.](https://justautomate.ai/klienci/surf-inc.svg)![SurfBot](https://justautomate.ai/agenci/surfbot-80.webp)SurfBottheir agent Surf Inc., a premium clothing brand with an online shop and two physical stores 16 people from the company work with the agent 21 systems wired in permanently 89 processes run without a human 323 thousand lines of the agent’s code A PrestaShop store wired into the order system, the ERP, the data warehouse and the ad platforms. The agent launched in March 2026 and was fully running by May. Sixteen people use it, warehouse and physical store staff included. 8 of 21 systems it works in - PrestaShop - BaseLinker - Comarch Optima - BigQuery - Google Ads - Meta Ads - GA4 - Zalando Before From February to August 2025 every unit of ad spend came back as 3.53 in online revenue, and advertising absorbed 28.3 percent of that revenue. After In the same months of 2026 the ratio is 4.19 and advertising is down to 23.9 percent of revenue. The ad budget is 6.9 percent lower with sales 10.3 percent higher. Return on ad spend from 3.53 to 4.19, 18.5 percent higher, on a budget 6.9 percent lower And in your shop ## What if our process looks different? Four sentences come up in almost every consultation. Here is what we answer to each of them. There is also a list of things we do not automate, and we go through it in the workshop. We do not hand a machine the decision to accept a complaint, or to release money before the goods have been checked. We do not build a model that sets prices on its own; prices move on a rule somebody on your side approved. And we do not automate a process nobody in the company can describe, because that produces faster mess, not less of it. - We have BaseLinker, it already does this. For orders, statuses and couriers it usually does, and we will tell you so. The gaps sit in the same places every time: partial shipments, handing the sales document to accounting, predicting a stock-out and running a return against a deadline. If all that is missing is configuration you have already paid for, we show you how to finish it. - Every second order here is an exception. Then we start by writing the exceptions down, not by automating anything. It usually turns out that five repeatable situations account for most of the traffic: one line missing, a pre-order, an address change after dispatch, a parcel nobody collected, a return sent back with no filing. The automation runs the standard path and routes the rest to a person with a deadline. - We are afraid the automation will send a customer nonsense. That is why the first stage runs next to your current process with customer messaging switched off. We compare its decisions against what a person did on the same orders from the past few weeks. It speaks to a customer only after the switchover, and the off switch stays on your side. - The warehouse will not click through a new panel. We are not adding a panel. The changes happen between systems, so the warehouse works on the same scanner and the same picking list, it just arrives ready. In our own shop the warehouse is one of the active users of the agent, not a separate rollout. Shop platforms Shopify, WooCommerce, PrestaShop, Magento Orders and marketplaces BaseLinker, Allegro, Amazon, eBay, Google Merchant Center Warehouse and couriers InPost, DPD, DHL, GLS, WMS systems, pick and pack Accounting and ERP Fakturownia, Xero, sevDesk, Comarch Optima, Subiekt Support and notifications Transactional email, Slack, Microsoft Teams, Google Workspace Automation layer Make, n8n, Zapier, custom API integrations ## Seven stages of the order path and what we connect at each one This is the road one order travels, broken into stages. Almost nobody takes all of them at once, so in the workshop we pick the stage where the most time or the most money is lost today. | Stage of the order path | What happens without a person | What we connect there | | ----------------------- | ------------------------------------------------------------------------------------------------------- | ---------------------------------------------- | | 01\. Order intake | stock reserved, picking document created and the customer confirmed, all from one event | platform, BaseLinker, warehouse system | | 02\. Picking | a pick and pack list with locations, a missing line flagged before packing starts | WMS, scanner, stock sheet | | 03\. Dispatch | the label built from order data, the tracking number on the platform and with the customer | InPost, DPD, DHL, GLS through the order system | | 04\. Sales document | the invoice or receipt is created after dispatch and reaches accounting with nothing retyped | Fakturownia, Xero, Comarch Optima, Subiekt | | 05\. Stock and listings | one stock figure for the shop and the marketplaces, the listing pulled at zero and restored on delivery | platform, warehouse, Allegro, Amazon | | 06\. Return | the return label, the status and a task with a deadline counted from the day of filing | return form, warehouse, payment system | | 07\. Complaint | a filing with a reason and a photo reaches a person together with the order and shipping history | ticket system, email, Slack | The order of the stages is natural, the order of the rollout does not have to be. We usually start with order intake and dispatch, because those two touch every parcel, which means every day. How we do it ## Five steps from retyping to a flow Every step ends with something you can judge before you decide on the next one. 1. ### Workshop and the list of exceptions One week. We walk the order path stage by stage and note every place where data is retyped by hand. Exceptions get their own list, because they decide the shape of the solution: stock-out, partial shipment, address change, return, complaint. 2. ### Choosing the first stage We take one stage, usually order intake or dispatch, and count it before we start: orders a day, hand-offs per order, mistakes a month. Without that number there is no way to say later whether the rollout worked. 3. ### Running it alongside The flow runs in parallel with your current process, with customer messaging switched off. We compare its decisions against the decisions a person made on your orders from the past few weeks, exceptions included. 4. ### Switchover with an off switch Once it checks out we switch it to production. An off switch stays on your side, along with an error queue that shows what did not go through and why. Someone in support learns to read it and close it, without raising a ticket with us. 5. ### The next stage and the measurement After 30 and 90 days we check the metrics agreed in the workshop: how many orders went through without a person, how many landed in exceptions, how many dispatch mistakes are left. Only then do we add the next stage, down the same track. Before and after ## How the day changes for the person running online sales ### Before - The morning starts by retyping the orders that came in over the weekend - Tracking numbers get pasted by hand, one order after another - A stock-out shows up at packing, after the customer has already paid - Returns sit in a box until somebody finds the time to enter them ### After - Weekend orders are waiting in the packing queue, not in an inbox - The tracking number is with the customer before anyone opens a panel - A missing line is flagged at intake, together with a partial shipment option - A return has a status, a deadline and an owner from the moment it is filed The first stage is visible on the day of the switchover. The share of orders handled without a person and the number of exceptions get measured after 30 and 90 days, on the metrics set in the workshop. For the person signing this off ## What the owner and the head of operations get out of it Growth in sales stops meaning a hiring round. The work between systems no longer grows with the number of parcels, so the question how many people do we need for November turns into a question about warehouse throughput. For the budget conversation you hold the count of orders handled without a person, not an impression that things feel better. The second thing is the project risk you will be judged on. Every stage goes into production next to the current process and with an off switch, so the worst case is going back to what you have today. The process map and the list of exceptions stay with you, including on the day we stop working together. - The process and its exceptions written down, no longer held by one person in support - Each stage launched next to the current one, with an off switch and an error queue - Holidays and staff turnover stop halting dispatch and returns - Measurement after 30 and 90 days, on metrics agreed before the start What it costs ## The workshop has a published price, the rollout is quoted after it The process workshop costs **EUR 3,500 net per week** and runs from 1 to 7 weeks. A shop usually needs one or two weeks, because the order path crosses several systems and each one has to be looked at from the inside. We quote the rollout after the workshop, once it is clear how many stages are in scope and what state the data in those systems is in. Tool costs you pay straight to the vendor: the order system, the automation layer such as [Make](https://www.make.com/en/help) or [n8n](https://docs.n8n.io/), and the language model, at EUR 0.005 to 0.07 per query at provider list prices. No margin from us and no invoices passing through us. When the automation layer runs on your own server, that cost disappears. - ### Process workshop EUR 3,500 net per week, usually one or two for a shop A self-contained piece of work. It ends with a map of the order path, a list of exceptions and a plan you can execute without us. - The order path broken into stages, with every point where data is retyped - The list of exceptions: stock-out, partial shipment, return, complaint - A tooling recommendation: configuration in the order system, Make, n8n or the API - The metrics to measure at 30 and 90 days Start with the workshop - Most popular ### Workshop and one stage in production Quoted after the workshop scope follows from the number of systems and their state The option most shops choose. The first stage in production 2 to 4 weeks after the workshop, tested on your own orders. - Everything in the workshop - The platform connected to the warehouse, the courier and accounting - Exception handling, an error queue and an off switch on your side - Launched next to the current process, switched over once it checks out - Training for the person who runs it on your side Talk about a rollout - ### Shop processes and an AI agent Quoted after the workshop For a shop where questions about orders, stock and returns come up a dozen or more times a day. - Everything in the stage rollout - BOND wired into the shop, the warehouse and the shipping system - Alerts on parcels stuck at the courier and on returns close to their deadline - Access for customer service and the warehouse, with no panel to log into Ask about the agent If the workshop shows that all you need is to finish the configuration you already have in the order system or on the platform, we say so and show you how. The workshop ends with a document that stands up without us, including in front of a different supplier. If there is nothing worth automating, we will say it plainly and we will not propose a rollout. ## Frequently asked questions Which stage of the shop do you start with? With the one that touches every order and eats the most time, so most often order intake or dispatch. We set the order in the workshop, based on how many orders come in a day and how many manual hand-offs sit on one of them. We do not take all seven stages at once, because then there is no way to say which one produced the effect. What does the path of one order look like after the rollout? A paid order reserves stock, creates the picking document and the confirmation for the customer. The warehouse packs from a finished list, the courier system gets the data with nothing retyped, and the tracking number goes back to the platform and to the customer. After dispatch the sales document is created and goes to accounting. A person steps in only where a decision is needed. What happens when an item in the order is out of stock? The stock-out shows up at intake rather than at packing, because reserving stock is the first step. The order then goes down the exception path: a partial shipment offer with a date for the rest, or a call to the customer. A person makes the decision, but it lands on their desk with a deadline instead of being discovered next to a cardboard box. How do you handle returns and the refund deadline? A return filed through the form creates the return label, sets the status and opens a task with a deadline counted from the day of filing, because that is the day the clock starts in consumer sales. The refund starts once the goods are booked into the warehouse or once there is proof of posting, whichever your policy allows. The transfer itself stays a human decision. What do you not automate in a shop? The decision to accept a complaint, releasing money before the goods have been checked, and the conversation with an angry customer. We also do not build a model that sets prices on its own; prices move on a rule somebody on your side approved and can switch off. And we do not automate a process nobody in the company can describe, because that produces faster mess. We have BaseLinker and platform modules, is that not enough? For orders, statuses and couriers it usually is, and we will tell you that in the consultation. The gaps appear at partial shipments, handing the sales document to accounting, predicting a stock-out and running a return against a deadline. That is where we add a layer in Make, n8n or through the API, and the order system stays exactly where it is. How long does one stage take to roll out? One week of workshop and 2 to 4 weeks for the first stage, counted from the end of it. For part of that time the flow runs next to your current process, so its decisions can be compared with the decisions a person made. The timeline grows when the data in your systems needs tidying, and we say that in the workshop rather than mid-rollout. Who watches this after the rollout and what if a flow stops? Every flow has monitoring and an error queue describing what did not go through and why, plus an off switch on your side. A stopped order stays in the queue with its data, so once the cause is fixed it moves on instead of vanishing. Someone in support reads and clears that queue after a short training, without raising a ticket with us. ## Let us walk the path of one of your orders together The free consultation takes 30 minutes. We take one order from the past week and go through it stage by stage: who retypes what, where it waits and what happens when it goes wrong. At the end we say which stage is worth taking first, and what you only need to finish configuring. You do not have to give us access to any system at this point. We run at most two rollouts at a time, so that someone is actually sitting on each of them. [Book a free consultation](https://justautomate.ai/en/contact/) **PS.** If somebody in your company spent this week pasting tracking numbers from one window into another, that is exactly this process. An order can travel from click to dispatch with nothing retyped by hand, and the people stay on the exceptions that genuinely need a decision. ## Related - [AI and automation for e-commerce](https://justautomate.ai/en/industries/e-commerce/) - [Automatic inventory update](https://justautomate.ai/en/automations/automatic-inventory-update/) - [Invoice automation](https://justautomate.ai/en/services/invoice-automation/) - [BaseLinker integration](https://justautomate.ai/en/integrations/baselinker/) - [n8n automation](https://justautomate.ai/en/services/n8n-automation/) - [BOND: the business AI agent](https://justautomate.ai/en/services/bond-business-ai-agent/) --- **JustAutomate** · contact: contact@justautomate.ai · form: https://justautomate.ai/en/contact/