Short answer: in an e-shop four jobs pay off most: customer messages about orders, returns, product entry and orders to suppliers. The threshold where it's worth starting is roughly 500-600 customer messages a month. And what you need to count isn't the number of orders, but how much time communicating with customers takes.
Four jobs that pay off most
- 1. Customer messages about orders. The biggest volume and the least interesting work. “Where's my parcel”, “when will I get it”, “do you still have this in my size”. The agent looks up the order in the system itself and answers with concrete information instead of pointing to the customer account.
- 2. Returns and exchanges. The process is always the same: check the deadline, explain the procedure, generate the document, register it. It's a recipe you can write down end to end, while leaving the refund approval to a person.
- 3. Product entry and descriptions. A supplier file arrives with hundreds of items that need to be sorted into categories, given descriptions and have their specifications tidied up. It hurts most when the same catalog is filled into two shops.
- 4. Orders to suppliers based on stock. The system sees what's running out and how fast it sells, and prepares the order. You only approve it. The benefit here isn't only hours, but also what you didn't sell when the product ran out.
In an e-shop, automation pays off twice: in hours saved and in orders you didn't lose because you answered within a minute.
Seasonality changes the whole math
I'm stressing this part because in e-commerce it matters more than in any other area.
You hire your team for the average, but customers write at the peak. In November and December the message flow can grow several times over, and that's exactly when replies run late, negative reviews appear and some buyers go to a competitor. Hiring a person for two months is impossible, and training them is even harder.
A system has no peak. It answers the same way at fifty messages a day and at five hundred. So in an e-shop, calculating the return on the average alone means undervaluing automation. The real value shows in the weeks when you otherwise simply wouldn't have kept up.
Where AI gets e-shops wrong
As everywhere, staying quiet about errors would be dishonest:
- Promises about delivery dates. Once the parcel is with the courier, nobody knows the exact date. The system has to say what the carrier shows, not comfort the buyer with its own guess.
- Product specifications. If the supplier's data is inaccurate, the agent will copy those inaccuracies faster than a person could notice. Review is essential wherever the specification decides the purchase.
- Angry customers. When a person is already disappointed, an automatic reply, even a correct one, adds fuel to the fire. Those conversations must go to a person straight away.
In practice this means a hybrid: the agent handles the repetitive majority, and flags a few percent of cases and passes them on with full context. I wrote about this split separately.
What the numbers look like
Take an e-shop that gets 900 customer messages a month by email and on social media, spending an average of five minutes on each, including the lookup in the system. That's 75 hours a month.
| By hand | With AI | |
|---|---|---|
| Time per month | 75 h | ~12 h for complex cases |
| Cost ($46.89/h + our price) | $3,517 | up to $750 |
| Difference per month | ~$2,767 | |
| Per year | ~$33,200 | |
Calculated on our model: up to $10 for each hour of AI work, against $46.89 an hour for a person, the average US private-industry employer cost per hour worked, wages plus benefits (BLS, June 2026). Run your own numbers in the calculator. And note: peaks and orders not lost aren't included here, and in reality they're often exactly what tips the scale.
When it's not worth it. If you get 100 messages a month, that's about 8 hours of work, worth roughly $375. After our $300 monthly minimum and the setup, there would be next to nothing left of the savings, and I say exactly that. Better to sort out the FAQ on your website and use a ready-made chat tool.
Integrations: the shop is the easiest part
Popular platforms (PrestaShop, WooCommerce, Shopify, Magento) have interfaces for both reading orders and performing actions. The shop itself is the least likely to cause problems.
The difficulties start around it: the warehouse software, accounting, courier services and the channels customers write through. In Lithuania the typical set is email, Facebook and Instagram messages, sometimes the phone, and a marketplace on top. Every channel and every system is a separate piece of work, and they're exactly what drives the price.
So before asking for an offer, have answers to two questions: which channels your customers write through, and which systems need to be connected.
Where to start
Start with one channel and one type of question, most often order status by email. It's the biggest volume, the clearest rules, and the result is visible within a couple of weeks.
Once that works, social media gets connected, then returns, then product entry. Trying everything at once before the season is the most common mistake: the peak will come before the system settles in.
And one timing tip: if you're aiming at the autumn season, you need to start in summer. Setup with a pilot takes weeks, not days.
Questions about AI in e-shops
What can AI automate in an e-shop?
Four jobs pay off most: customer messages about order status, handling returns and exchanges, product entry and descriptions, and orders to suppliers based on stock levels. All of them repeat every day and have clear rules.
Can AI answer customers instead of me?
Repetitive questions, yes. Those are mostly questions about order status, delivery times and the return procedure. Angry customers, negotiations over compensation and unusual cases must be passed to a person with full context, not handled automatically.
At what order volume does automation pay off for an e-shop?
What you need to count isn't orders, but customer messages and the time spent on them. The rough threshold is 500-600 messages a month if each takes a few minutes. For smaller volumes, a ready-made chat tool or a proper FAQ on the website is usually cheaper.
Does AI work with PrestaShop, WooCommerce or Shopify?
Yes. Popular e-commerce platforms have interfaces through which you can both read orders and perform actions. The harder part is usually not the shop itself but the systems around it: the warehouse, accounting and courier services.