• Written by Azhar Aliyar, Director

When a customer service representative checks the order history in the customer relationship management system, they may find that the product the customer is querying is no longer listed in the current product catalogue, even though the customer's order was fulfilled just a few weeks ago. This discrepancy can cause confusion and delays in resolving the customer's issue, and it is an example of the challenges of integrating AI systems with the systems of record in e-commerce and retail, where fast-moving assortments and direct customer signal can make it difficult to keep data consistent across different systems.

Challenges of e-commerce and retail integration

The e-commerce and retail industries pose unique challenges for AI systems integration, particularly when it comes to assortment decisions, returns and their root causes, and promotion effectiveness. Assortment decisions, for example, involve selecting the right products to offer to customers at the right time, which can be a complex task given the constantly changing product offerings and customer preferences. Returns and their root causes can also be difficult to track and analyze, as they may involve multiple systems and stakeholders. Promotion effectiveness is another area where integration can be challenging, as it requires tracking the impact of various promotional campaigns on sales and customer behavior.

Most people get wrong the idea that integrating AI systems with e-commerce and retail systems is a straightforward process that can be accomplished with minimal effort and resources. However, the reality is that it requires careful planning, execution, and ongoing maintenance to ensure that the integration is successful and provides accurate and consistent data.

Read paths, write paths and the need for care

When integrating AI systems with e-commerce and retail systems, it is essential to consider both read paths and write paths. Read paths refer to the process of retrieving data from the systems of record, while write paths refer to the process of updating or inserting data into these systems. While read paths are generally straightforward, write paths require more care and attention, as they can potentially introduce inconsistencies or errors into the data.

A shift supervisor checking the maintenance log, for example, may need to update the log with new information, which requires a write path that is carefully designed to ensure data consistency and accuracy. If the write path is not properly implemented, it can lead to errors or inconsistencies that can have significant consequences, such as delays or lost sales.

Reconciling definitions that differ between systems

One of the significant challenges of integrating AI systems with e-commerce and retail systems is reconciling definitions that differ between systems. For example, the definition of a "customer" may vary between the customer relationship management system and the enterprise resource planning system. Similarly, the definition of a "product" may differ between the product information management system and the inventory management system.

To reconcile these differences, it is essential to establish a common set of definitions and standards that can be used across all systems. This may involve creating a data governance framework that outlines the rules and procedures for data management and integration. By establishing a common set of definitions and standards, organizations can ensure that their data is consistent and accurate, and that it can be easily shared and used across different systems.

Latency budgets and precomputation

When integrating AI systems with e-commerce and retail systems, it is also essential to consider latency budgets and precomputation. Latency budgets refer to the maximum amount of time that can be spent on processing and retrieving data, while precomputation refers to the process of preparing data in advance to reduce processing time. By carefully managing latency budgets and precomputing data, organizations can ensure that their AI systems can provide fast and accurate responses to customer queries and other requests.

For example, a category manager comparing two suppliers may need to retrieve data on their pricing and delivery schedules. By precomputing this data and storing it in a cache, the category manager can quickly retrieve the information they need without having to wait for the data to be processed and retrieved from the underlying systems. To learn more about how AsscherAi can help with latency budgets and precomputation, visit our website at www.asscher.ai.

Limitations and wrong choices

While integrating AI systems with e-commerce and retail systems can provide significant benefits, there are also limitations and wrong choices to consider. For example, AI systems may not be suitable for situations where data is highly unstructured or where the decision-making process is highly complex and nuanced. In such cases, other approaches, such as manual analysis or expert judgment, may be more appropriate.

Additionally, organizations should be cautious not to over-rely on AI systems, as they can introduce new risks and challenges, such as data bias and errors. To get in touch with our experts and discuss how AsscherAi can help with your e-commerce and retail integration needs, visit our contact us page.