When sellers think about automating a marketplace store, they usually picture sales-analytics services and tools for managing prices on the storefront. But there is a second half of the job those tools never touch: purchasing. Supplier emails with price lists and invoices, recalculating margin after every change in purchase price, payment deadlines kept in your head or in a phone note. This is the routine that eats your evenings — and this is exactly what an AI operator handles. In this article we break down, mechanic by mechanic, what it does for a seller and what honestly stays outside its reach.
What the purchasing routine actually involves
A typical marketplace seller's day splits into two parts. The first is the storefront: listings, prices, ads, reviews. For that there are seller dashboards and dozens of services. The second is supply — and here there are almost no tools.
Supply lives in email and messengers. A supplier sends a new price list: you have to open it, compare it with the old one, work out which lines went up and whether margin has slipped into the red. An invoice arrives: you must not forget to pay it before the deadline, or the next shipment gets held up. Someone promised to confirm stock and went quiet: you have to remember and chase them. Multiply that by a few dozen suppliers and you get a second job hidden inside the first.
The problem is not that any single step is hard — it is their number and how scattered they are: some in email, some in a spreadsheet, some in your head. An AI operator — HORUVIA, for example — pulls this flow into a single loop and works through it like an attentive purchasing assistant.
How store automation works on the purchasing side
The AI operator connects to your work email over IMAP and starts with the most time-consuming task of all — sorting the inbox.
Sorting supplier email
Every message is classified: which supplier it is from and what is inside — a price list, an invoice, an order confirmation, stock information, or just correspondence. Attachments are filed by supplier and by document type. Instead of a wall of a hundred unread emails, the seller sees structure: here are the new price lists from this week, here are the invoices awaiting payment, here are the messages nobody has answered.
A memory of purchase prices
From price lists and invoices, the system builds up a price history for every line from every supplier. This is purchase accounting for a marketplace business, only without manual entry: prices land in memory from the documents that arrive in your inbox anyway. When a supplier sends an updated price list, you immediately see what went up and by how much — so you can recalculate margin on specific lines before the purchase eats into profit. This memory is built for large catalogs with dozens of brands and thousands of items, but it is just as useful for a small store: the mechanic is the same. For more on how this works, see our article on tracking purchase prices.
The invoice as a payment draft
When an invoice arrives, the system reads it — including scans, where OCR does the work — and prepares a payment draft. One detail matters: the amounts and payment details are handled by deterministic code, not a language model, so there is no room for "hallucinations." And no payment ever goes out on its own: a person always confirms the draft. The AI removes the manual copying of figures, but the decision about money stays with the owner.
Telegram alerts: an inbox that reminds you about itself
A seller does not need to sit inside an interface to avoid missing things. Key events arrive in Telegram:
- an invoice came in — with the amount and the supplier;
- an email carries a time-limited offer — for example, a special price until the end of the week;
- a message to a supplier has gone unanswered for too long;
- a certificate or contract is coming up on its expiry date;
- something in the pipeline went down — and a separate message when it comes back up.
The logic is simple: the system stays quiet while everything runs normally and speaks up when a decision or a person's attention is needed. The same reminder mechanism works in the other direction too — for money that is owed to you: we covered how that works in our piece on payment reminders.
The deadline engine: payments, contracts, shelf life
A separate loop watches the dates. Into it go payment due dates on invoices, contract end dates with suppliers, conformity certificates, and — for product categories with a limited shelf life — the expiry dates of batches. For each date the system reminds you ahead of time and in several waves — a couple of months out, a month out, and a week out. For a seller this means a certificate will not expire "out of nowhere" in the middle of peak season, and a supplier payment will not fall through because an invoice got buried in the inbox. You do not have to enter the dates one by one by hand: the engine picks them up from the same documents the inbox has already sorted.
The same loop for every marketplace, with its own emphasis
The purchasing side is built the same way across marketplaces: the same suppliers, price lists, invoices, and deadlines. So there is no separate product per platform — it is the same "retail" onboarding preset, which includes email sorting, price memory, and alerts configured for retail.
The emphasis shifts in planning. For some marketplaces it is more critical to assemble shipments ahead of time, and here the order recommendation helps: it works out demand from seasonality, the product's ABC classification, and current stock on hand. This is not a "magic button" but a calculation on a clear formula that saves hours of work in spreadsheets and gives you a starting number to decide from — the final shipment volume is still approved by the seller.
Two tools hold the overall picture together: Company Pulse, where orders to suppliers are visible by stage — from request to arrival at the warehouse — and a kanban task board that makes clear what is currently in progress for you and your team.
What the AI operator does not do: an honest boundary
Here it matters not to overpromise. HORUVIA does not run your marketplace storefront:
- it does not create or edit product listings;
- it does not manage ad bids or promotion participation;
- it does not reply to customer reviews;
- it does not handle storefront repricing.
Those tasks belong to the marketplace dashboards and specialized services — and it makes sense to keep using them. The AI operator covers the other half of the business: supply, documents, money, and deadlines. These two halves complement each other rather than compete, and it is more honest to say so plainly than to paint a picture of "one platform for everything."
There is another boundary — money. The system prepares payment drafts but does not execute them: confirmation always stays with a person. This is a deliberate architectural choice, not a temporary limitation that will "soon be removed."
What launch looks like for a retail seller
Marketplace automation often stalls at rollout: a seller has no weeks to spare on setup and training. That is why launch is built on onboarding presets — ready-made configurations per segment. The retail preset already includes the usual email categories, alert templates, and a memory structure for purchase prices.
In practice it looks like this: email connects over IMAP, and if needed a small agent is installed on the work computer to send documents from a chosen folder to the server — and the system starts sorting incoming mail and building up memory. If your bookkeeping runs in accounting software, there is a basic data exchange. The first few days go into letting the system see your real suppliers and documents; after that it works off the accumulated memory, and every new price list or invoice falls into a structure it already knows. We laid out a realistic day-by-day launch plan in our article on automation in a week.
History is worth a separate note. If you have archives of price lists and invoices piled up — in email or in folders on disk — you can hand them to the system: OCR reads the scans, and the price memory fills up not from zero but straight from your real purchasing history.
Where to start
If your store's supply lives in email and a spreadsheet, start with a simple measurement: count how many hours a week go into sorting mail, checking price lists, and tracking payments — there is a time-savings calculator for exactly this. And to see how the AI operator would sort your own inbox and which alerts the retail preset would set up, request a demo on the HORUVIA homepage — we will show it on real scenarios, without marketing slides. Answers to common questions — from the security of email access to working with accounting software — are collected in the FAQ.