Most ecommerce sales close in the inbox, not on the product page. "Do you have this in a large?", "when will it ship?", "where is my order?" — while those questions go unanswered, carts stay empty and full carts never reach checkout. The E-Commerce Sales Agent is a digital worker that takes those conversations over: it knows your catalog, queries your orders, and carries the customer through to purchase.
Why order and stock questions pile up in your inbox
Probably familiar:
- Customer questions piling up in the marketplace seller panel, messages waiting on WhatsApp, an unread Instagram DM folder — three places, the same questions.
- Answering "do you have it in black, size L?" means opening the stock sheet first.
- "Where is my order?" is the most repeated message of the day, and every one of them means typing an order number into a panel.
- Nobody follows up with the customer who added to cart and left.
- The purchase-intent message that arrives at 9pm gets answered the next morning — that customer already bought elsewhere.
- Answering "how much am I getting paid out this week?" means logging into the seller panel and downloading a report.
The problem is not neglect. It is that all of this is done manually, at once, and only during office hours.
What an AI sales assistant for ecommerce takes over
The agent picks up the following:
- Answers product, size, color and stock questions from real catalog data
- Recommends products, offers alternatives, cross-sells
- Queries order status and shipment tracking
- Queries your Trendyol orders and order details
- Queries Trendyol claim and return records and reports their status back to the customer
- Queries seller-side settlements, payment line items and cargo invoice items
- Follows up proactively with customers who abandoned their cart, using a ready trigger template
- Records return and exchange requests and routes them to your team
- Hands the conversation to live support when it cannot resolve it
Product search is semantic, not keyword matching: when a customer writes "I'm looking for a thick, dark, winter sweater," the agent finds the right items in your catalog.
A WhatsApp order tracking bot in action
A customer writes in:
Customer: Where is my order? My order number is 4821.
Agent: Your order was handed to the carrier today. Estimated delivery is tomorrow.
Customer: Can you send the tracking code?
Agent: Of course. Your tracking code is ready, and you can also see it in your account.
Customer: By the way, do you have that same item in a large?
Agent: Large is in stock, in black and anthracite. Would you like it?
Three jobs done in one conversation: order queried, tracking shared, a new sale opened. No panel was opened for any of it.
You stay in control of your Trendyol and catalog data
The agent works inside the boundaries you set:
- You choose the tools. You decide which systems the agent can reach. Every tool has three modes: run automatically, ask for approval before running, or stay fully disabled.
- SMS verification on personal data. Tools that return personal data such as order details can be placed behind a one-time SMS code.
- Every tool call is recorded. Each tool the agent runs — with the parameters it was called with and what it returned — is written to the audit log in your panel. Those records cannot be edited afterwards, and sensitive fields are masked before storage.
- Every conversation is visible. Whatever channel it came from, each conversation is listed and readable in the panel.
- You can stop it. You can deactivate the agent at any time, disable a single tool, or take the conversation over through live support.
How it runs Trendyol order and finance queries
Briefly, on the technical side: product and stock data come from Smart Catalog through semantic search. Shipping terms, return policy and store rules are read from the Knowledge Base. Trendyol order, claim and finance queries run in real time inside the conversation through the MCP tools in the Integrations layer. Self-starting flows such as abandoned-cart reminders are set up with the trigger templates in Smart Automation. Multi-step requests are executed in order by Task Orchestration.
The e-commerce sales agent's primary metric
Number of closed sales and revenue. The agent is measured by the sales it closes, not the messages it answers.
Recommended channels: WhatsApp, Instagram DM, Web Widget
Other digital workers
This agent is part of our digital workforce. The two closest to e-commerce:
- Customer Support Agent — takes over post-purchase questions and the support load.
You can put an ecommerce sales agent to work today. Pick the persona, connect your catalog, open your channel.