1. Implementation model: real-time, scheduled or hybrid model?

Interfaces to Sonet CGI ERP are always designed based on customer needs. We do not offer a single fixed interface that the customer should adapt to, but the ERP connections built on AI Commerce platform are highly configurable and scalable. Since the same technology is used by several different merchants, maintenance costs are shared among the tenants, and this achieves better operational reliability, service reliability and sufficiency of resources for further development compared to interfaces that are completely customer-specific.

Sonet connections are implemented using a GraphQL interface for data retrieval (queries from Sonet), and a REST-based XML format (send to Sonet). The solution enables flexible data structure querying ( GraphQL ) and standardized, error-reducing XML messages (send).

In addition, messages move through a procession , where they are stored and released to the receiving system within its resource and request-throttling constraints. This allows us to handle very large order volumes without congesting the system (e.g. 100 e-commerce orders per second). With the default configuration, messages sent to Sonet are released by a process running at minute intervals . This model is a hybrid model: it combines real-time storage and scheduled transmissions. If errors occur, error messages are stored in the system and can be resent if necessary.

For messages coming from Sonet, the system calls Sonet with the default configuration every minute and retrieves any new or changed data using the DateModified timestamp. This keeps the data transferred to a minimum (changes only) and keeps synchronization consistent.

Scheduling is handled by AWS EventBridge scheduling service and AWS Lambda functions, ensuring that these cron jobs are not dependent on local servers. At the same time, the serverless environment facilitates high availability and reliable scaling.

2. Additional variant information (e.g. “Boot size = 43”)

In the product model implementation, AI Commerce supports several different product types (e.g. simple , configurable , variant, option, group collection, etc. ). Different sizes and other product variations can be handled by linking each variant to the corresponding item in Sonet.

The most common practice is that each variant is its own product line in Sonet, allowing its balance to be managed separately.

If ERP does not have separate items for each variant, additional information about the variants can be passed to the ERP's comment/additional information field . In this case, additional information, such as sizes or other properties, will be included in the order line and displayed in Sonet in the desired format.

3. B2B customer segments, roles, rights and login

Customer segments mentioned in the RFP needs document (such as B2B and reseller customers) may already be pre-created in Sonet , in which case these customers do not need to create new accounts separately. In the AI Commerce platform:

It is possible to synchronize customer IDs and segment information directly from Sonet.

You can limit roles and rights (who can see prices, who can place orders) with flexible rules.

The login process can be implemented so that the customer account defined in the ERP serves as the online store ID, or customers are identified from the ERP before an account is created for them on AI Commerce side.

Since each company has its own practices, these implementations are made according to your needs . In practice, AI Commerce core development is responsible for customizing the solution, either as a general-purpose implementation directly into the core code, or by partners in a serverless environment without changing the core code.