You automate customer service using order data by connecting the chatbot to your shop and ERP. The chatbot then retrieves the status of an order during the conversation, registers returns, and creates tickets. Before doing so, the customer verifies their identity in a standardized step within the dialogue. The chatbot transfers cases that require a decision to your team, complete with the full conversation history. This enables it to resolve shipping status, delivery times, returns, and complaints automatically around the clock.

This article explains the mechanics behind it. An overview of all recurring service use cases can be found in 15 tasks for the chatbot.

Which service cases does the chatbot resolve autonomously using your order data?

The chatbot takes over common inquiries in e-commerce as soon as it has access to orders, inventory, and product data. The table shows the required data and the chatbot's task for each use case.

The chatbot only asks for identification details when addressing a specific order. The E-Commerce page shows how the chatbot provides consultation in the online shop.

How do customers view the current status of their order in the chat?

The chatbot accesses your systems via interfaces while the conversation is ongoing. It reads data from them and writes back to them as well. For example, it queries a status, registers a return, or creates a ticket. The dialogue then responds based on the system's output. For instance, the chatbot offers to initiate a return if the ERP marks the order as delivered.

Mercury.ai provides standard connectors for Shopware, Shopify, and Magento. On the ERP side, Mercury.ai integrates with SAP and Microsoft Dynamics 365. All other systems can be connected via the open REST API or custom webhooks.

A standard connector is ready within a few days. A custom integration takes two to four weeks. Your team provides access once, approves the data fields, and signs off on the implementation. Mercury.ai handles the integration, mapping, monitoring, and maintenance during API updates. The Integrations page lists all systems with pre-built connectors. The guide Connecting an AI Chatbot to Shopware describes the Shopware integration in detail.

How do you ensure that only the purchaser sees their data?

The chatbot must know who is asking. Only then does it disclose details about an order. It recognizes customers who are logged into the shop. Guests verify their identity in a standardized step within the dialogue. Your team determines which details are required. For example, this can be the order number and the email address associated with the order.

Mercury.ai uses an editorial dialogue for this step. Your team defines the workflow and copy once in the No-Code Studio. The dialogue then runs identically for every customer and follows predefined rules.

The chatbot validates every input before sending it to your system. For example, it detects an email address missing an @ symbol and prompts the user again. It does not disclose any data if the details do not match the order. Additionally, Mercury Intelligence verifies before each response whether the user is authorized to view the information.

How does a customer resolve their return in a single conversation?

A customer wants to return a jacket and provides the order number. In a standardized dialogue step, the chatbot asks for the email address and validates it. It then reads the order in the ERP and verifies that it has been delivered. It registers the return in the shop. The customer automatically receives a confirmation by email.

Your team does not need to take any action in this case. The return appears in the system just like any other.

What does your team receive when the chatbot hands over a conversation?

Some cases require human intervention. These include goodwill decisions, damages with an unclear cause, and upset customers. The chatbot transfers these conversations to Zendesk, Userlike, or the Agent Desk from Mercury.ai. Your team sees the full conversation history and all the information the chatbot has already collected. No one needs to ask the customer for their order number a second time.

The Agent Desk routes conversations by department, language, or workload. After the live chat, the conversation returns to a designated point in the chatbot. Outside of service hours, the chatbot creates a ticket. Your team can then continue working the next day with all the details readily available. The article on Availability outside of service hours explains this scenario in more detail. The guide on Selecting AI-powered live chat software outlines eight evaluation criteria.

How can you prevent returns right from the consultation phase?

Many returns originate before the purchase is even made. The customer might select the wrong dimensions or a variant that does not suit their needs. Consultation is therefore particularly critical for products where an incorrect recommendation leads to an expensive return.

The chatbot only recommends items that match the customer's specifications. To do this, it filters precisely by data fields such as dimensions, material, or compatibility. It asks a clarifying question if any information is missing. Customers describe their requirements in their own words, and the chatbot translates them into product attributes. It only answers using information from your sources. The system flags a gap if an attribute is missing from the product data, allowing your team to supplement the information.

TeeGschwendner uses this approach to advise customers in their Shopware shop until the right tea is in the shopping cart.

How is order data protected in the chat?

Mercury.ai processes all data in AWS Frankfurt, and therefore exclusively in Germany. Each customer's data is stored separately, and no data is transferred to third countries. Mercury.ai hosts and operates the language models itself and does not call external AI providers. Mercury.ai does not train any models using your customer data.

Your team defines how long chat histories are retained on a per-project basis. The retention period ranges from immediate deletion after input to the end of the contract. You can also store chat histories without user identifiers and apply a short retention period. Mercury.ai signs a Data Processing Agreement with you in accordance with Art. 28 GDPR. The article Where customer data flows in an AI chatbot details the journey of an inquiry across all systems.

Frequently asked questions

Does this also work via WhatsApp and Instagram?

Yes. The chatbot handles order inquiries in the chat widget, via WhatsApp, Instagram, and Facebook Messenger. You maintain knowledge, dialogues, and integrations only once across all channels. The standardized identification step runs identically on every channel.

What happens if the system cannot find the order?

The chatbot responds based on your system's feedback. It asks for the order number again or transfers the conversation to your team. Your team defines this workflow in advance in the No-Code Studio.

Does this also work in B2B shops with custom pricing and terms?

Yes. The chatbot identifies logged-in B2B customers and provides their customer-specific prices. The terms and conditions are retrieved live from the shop or ERP. Customers can also reorder within the chat, for instance using a list of SKU numbers or a CSV file.

Do we need to prepare our data beforehand?

No. You do not need to build a new database. The chatbot starts with the data you already manage today in your shop, PIM, and ERP. Mercury.ai handles the mapping and enriches the data until the chatbot can work effectively with it.

How long does implementation take?

Your chatbot typically goes live after four to six weeks. During the first two weeks, we review your data sources and define the three most important use cases. After that, we build the dialogues, integrate the shop and ERP, and launch with monitoring in place.

Would you like to know which service use cases your shop can automate first? Let's discuss it in a 20-minute call. Talk to us.