What stays with people
Approving copy and critical product data, large or unusual price changes, returns, and every customer who needs human support.
From manually written descriptions and Excel price lists to a catalogue that is updated automatically, with rules and approval where needed.
A store with a physical shop and an e-shop, 1,200 SKUs from 15 suppliers and listings on Skroutz and Google Shopping. The e-shop is run by two people, who also handle customer service.
Using Praion AI's Product Manager, it creates a draft description, FAQs and structured product fields for each SKU, based on the real catalogue data. Every piece of copy and every critical field is reviewed before publication.
It reads the supplier's file and, through the available integrations or feeds, prepares or applies updates to prices and stock in the e-shop and in Skroutz and Google Shopping listings. Changes above the store's approval threshold, unusual variations or products without a reliable SKU match are held for review.
It answers questions about an order's progress using only the actual data from the e-shop and the courier. If there is no clear update, or the question concerns a return, complaint or another exception, it is handed over to a person.
Approving copy and critical product data, large or unusual price changes, returns, and every customer who needs human support.
Catalogue maintenance stops consuming such a large part of the team's day. The two people running the e-shop can spend more time on promotions, partnerships and customers who need human support.
At a store like this one, the first step is the study. We talk to the people running the e-shop and map how a new product is added today, how price lists arrive and are entered, and which customer messages keep coming back. That shows which agent to start with.
The store defines the threshold above which a price change needs approval, which product fields count as critical and who approves the copy. It also defines which customer cases always go to a person.
Before they take on any work, the agents are tested with real price lists, products and messages from the store. When something does not fit their rules, they stop and flag it instead of guessing. After handover, we monitor how they work and adjust them when something changes, such as a new supplier or a new product category.