An online store can apply AI today across five processes: 24/7 customer support with its own assistant, abandoned cart recovery, product page creation, stock forecasting, and connecting orders to invoicing and the CRM. The potential we estimate at Zen Praxis: up to 35 hours a month recovered per person.
In a small online store the sale is almost never lost on the product. It's lost at 11:47 pm, when someone is torn between two sizes and there's nobody to answer. It's lost in the cart left half-finished while the customer compares in another tab. It's lost in the three Sunday hours you spend writing product descriptions instead of selling. And it's lost in the daily copy-paste between the store, the stock sheet and the invoicing software.
Your customers already shop with AI
The sixth Veepee–IESE study on the future of e-commerce in Spain —run by Kantar with 805 online shoppers— puts a number on something already in the air: 63 % of Spanish consumers use AI tools in their digital purchases. The question is no longer whether your customers use AI to buy. It's which side of the counter yours is on.
And there the study leaves the figure that matters most to a small store: 52 % trust AI assistants built into the store itself more than general-purpose tools like ChatGPT, which sit at 44 %. Proximity still wins, digital included. You don't need to compete with a general-purpose AI: you need to answer well, fast and with your own judgement in your own house.
The rest of the numbers describe your customer rather well: only 8 % would hand the buying decision over to an AI completely and 51 % want to understand why a product is being recommended to them —they want help, not an automated salesperson pushing, 30 % would pay for a specialised shopping assistant; 67 % browse the store for inspiration with no immediate intention to buy, and 63 % use the cart as a temporary comparison list. With online sales forecast to grow 7.4 % in 2026, the room for doing nothing keeps shrinking.
Five processes, in order of difficulty.
1. 24/7 customer support with your own assistant
The pain: the same questions, every day, almost always outside working hours. Which size should I order, how long does shipping take, will it arrive before Saturday, how do returns work, where is my order. Each one is small; together they're your afternoon. And the one that lands at eleven at night and waits until tomorrow usually ends up buying somewhere else, because an unanswered doubt is the number one reason a cart is left half-finished.
What gets automated: an assistant in your store that answers with your real information —your sizes, your delivery times, your returns policy, that customer's order status— in your brand's voice and at any hour. And that, as soon as the query has nuance or the customer is upset, hands it to a person without making them repeat anything.
What changes: you stop losing sales to night-time silence and stop living in the inbox. It's also the process where the 52 % figure works in your favour: shoppers trust your store's assistant more than an outside AI, as long as it's transparent and escalates to a human when it should. Difficulty: low-medium.
2. Abandoned carts and order updates
The pain: an abandoned cart reads like a failure and hardly ever is one. With 67 % of shoppers browsing for inspiration and 63 % using the cart to compare, that half-finished cart isn't a "no": it's a "not yet". The problem is that almost no small store does anything with it. And on the other side, "where's my order?" fills your inbox with messages that sell nothing.
What gets automated: a useful reminder in your own voice a few hours later —the product they left, the size they were looking at, the answer to the doubt that probably stopped them— plus a sequence of order status updates (confirmed, shipped, out for delivery, delivered) by email or WhatsApp, triggered by the shipment status itself.
What changes: you recover sales that were already half decided, and half of your customer support disappears, because nobody asks where something is when you've already told them. It's the automation on this list with the fastest and most measurable return. Difficulty: low.
3. Product pages and content
The pain: every new reference needs a description, variants, measurements, materials and, if you sell abroad, the same page in another language. Writing it properly takes time you don't have, so it ends up as four lines copied from the supplier —the same ones ten other stores have, so you neither convince the reader nor show up in searches—. And reviews go unanswered for exactly the same reason.
What gets automated: a draft product page generated from your data and your photos —description, features, variants, translation— that you review and publish. AI writes; the judgement stays yours, and that part isn't negotiable: a page with an invented detail is a return waiting to happen. Same with review replies: automatic draft, your review.
What changes: you go from writing to reviewing. Products reach the catalogue sooner and better described, which is exactly what makes people find them. Difficulty: low-medium.
4. Stock and demand forecasting
The pain: two ways of losing money at once. Running out of your best seller in the one good week, and having a warehouse full of what hasn't moved since March. You already have the information to avoid it —it's in your orders— but it lives in a spreadsheet export you open when it's already too late.
What gets automated: a system that watches your own sales and warns you in advance: what's about to run out at the pace of the last few weeks, what to reorder now given your supplier's lead time, what has been sitting still too long and needs promoting. Instead of a report nobody opens, an alert when there's a decision to make. It's the move from scattered spreadsheets to clear decisions, applied to your warehouse.
What changes: fewer stockouts in your best weeks and less money tied up on shelves. Difficulty: medium.
5. A connected back office: orders, invoicing and CRM
The pain: the order comes into the store, you note it in a sheet, move it to the invoicing software, copy the address onto the shipping label and, if the customer is a repeat one, look them up by hand in your inbox. None of that is selling: it's copy-paste. And it's where the errors creep in that later cost you a return or a wrongly issued invoice.
What gets automated: the integrations between the pieces you already use, so the order travels on its own from the store to the invoice, the label and the customer record. When off-the-shelf tools don't reach —because your flow has its own twist— it gets built to measure: that's what we did with Odonis in the dental sector, where clinic and lab share the status of every job in real time and the "did you get it?" calls disappear. The logic is identical in a store.
What changes: the manual shuffling of data between programs ends, and the errors it brought with it end too. It's the least glamorous process and the one that gives back the most hours by month end. Difficulty: medium.
What about appearing in ChatGPT when someone asks for recommendations?
That's the other half of the board, and the two are worth keeping apart. These five processes are about what happens inside your store; getting an assistant to recommend you when someone asks "where do I buy X?" is about what's published outside. They're complementary: outside visibility brings the customer in, and the assistant inside closes the sale —and 52 % versus 44 % says exactly that, trust is earned in your own house—. If the outside half interests you, we cover it in full in how to get ChatGPT to recommend your business.
Where to start without a technical team
Not all five. One. And in a small online store the right answer is usually between 1 and 2: customer support if you live interrupted and doubts are costing you sales; carts and order updates if what hurts is conversion and an inbox full of "where's my order?". If you're unsure, start with 2: it's the easiest to set up and the easiest to measure. And if you want to warm up with something simpler still, the list of 5 tasks you can automate this week works for any business.
The rule: measure sales and hours, not messages sent. Before you start, write down how many queries you get a week, how many carts are left half-finished, and how much time goes into product pages and copying data between programs. That number is your starting point and your proof that it worked.
You don't need to know technology or change platform: that's how we work at Zen Praxis —a free diagnosis to see which process is costing you most, a proposal with fixed timelines and budget, implementation without stopping the store, and training for the team—. The first solution can be running in weeks, with the potential to free up to 35 hours a month per person and the investment recovered in under six months.
How many sales slip away each week between unanswered doubts and half-finished carts? Book a free 30-minute call and we'll tell you what to automate first in your store, no jargon and no strings attached.