Our work

Commerce · AI automation

Three stores. One clear decision workspace.

AI-assisted price monitoring, stock checks and goods receipt preparation for a retail business.

For a merchant running several online stores, prices and stock data live in different systems. The ERP holds purchasing prices, the stores hold selling prices and stock, and wholesalers hold current offers and availability. We built a shared Hungarian-language workspace to compare them.

  1. 01 · The problemScattered data. Repeated checks.

    Prices, stock and supplier data had to be compared across three stores.

  2. 02 · The solutionEvery discrepancy in one place.

    The system brings data together, flags discrepancies and prepares reasoned proposals.

  3. 03 · ApprovalThe merchant decides.

    Current and proposed values appear side by side. Changes require human approval.

The problem: checking the same data again and again

The project connects three Shoprenter stores, a Tharanis ERP system and three suppliers. Purchasing and selling data must also be compared with prices from one designated competitor.

The suppliers had no usable API. Products had to be found, matched and checked on their websites. Different pack sizes or uncertain matches can distort a price comparison, so verifying product identity became the foundation of the process.

The solution: one workspace, clear proposals

We built a solution around the existing systems. Tharanis and Shoprenter data can be retrieved through their APIs; browser-based collection supports checks on supplier websites.

Verified product URLs are stored in a shared database, allowing the next check to begin from a known product page. Missing and uncertain matches remain visible and are never automatically marked as correct.

The dashboard brings available purchase, selling and competitor prices together with stock and availability in one product row. The decision list focuses on stock discrepancies and price differences above the agreed threshold.

The merchant stays in control

A dedicated approval screen shows the current value, proposed value and reason for the change together. Each proposal can be checked before the merchant decides.

The checking cycle was prepared as six daily batches followed by a report on day seven. Task progress, blocked work and open questions can be followed in one place.

From invoice to product bundle

For goods receipt preparation, we built PDF upload and a persistent processing queue. Net unit prices can be edited on the itemised invoice; euro invoices can use an invoice-specific exchange rate. Changing a price draft requires a new approval.

The system also considers bundle compositions in Tharanis. When a matched invoice line is part of a bundle, it identifies the affected bundles. A verified increase in purchasing cost produces a separate bundle-price review proposal, accounting for the quantity of that component in each bundle.

Rollout status

Built. Reviewable. Prepared for rollout.

Delivered so far

The project has delivered a working internal dashboard, verified API access, a persistent product-link database, an invoice intake queue and tested approval preparation. Data, discrepancies and decision rationales now share one workspace.

The next phase

The system is in a controlled rollout. Activating regular weekly runs and executing live price, stock and goods receipt changes belong to the next phase. Time savings and business impact will be measured during regular use.

Does one decision mean checking several systems?

We identify which checks are worth automating and where human approval should stay.

Let’s talk about your workflows

Next story

Mayo Chix