• Written by Azhar Aliyar, Director

A category manager for a fashion retailer checks the purchase order table in the enterprise resource planning system to confirm the delivery dates for the upcoming seasonal range, but the information does not match the data in the product lifecycle management system. This discrepancy is a common problem in textiles and apparel, where the complexity of range planning, supplier fragmentation, and markdown exposure make integration with AI systems a challenging task.

Complexity of textiles and apparel

The textiles and apparel industry is characterised by long lead times, seasonal ranges, and high levels of uncertainty, making it harder to integrate with AI systems. Range planning, for example, involves forecasting demand for products that may not be designed or sourced yet, and making decisions about production quantities and delivery dates. This uncertainty makes it difficult to ensure that the data in the AI system is accurate and up-to-date.

Supplier fragmentation is another challenge, as textiles and apparel companies often work with multiple suppliers, each with their own systems and processes. This can lead to inconsistencies in data formats and definitions, making it harder to integrate the data into a single AI system. Markdown exposure is also a concern, as textiles and apparel companies need to be able to quickly respond to changes in demand and adjust their pricing and inventory levels accordingly.

Read paths, write paths and data consistency

When integrating with AI systems, it is essential to consider both the read paths and write paths of the data. The read path refers to how the AI system accesses the data from the systems of record, such as the enterprise resource planning system or the product lifecycle management system. The write path, on the other hand, refers to how the AI system updates the data in these systems. The write path requires more care, as it is critical to ensure that the data is handled correctly and consistently across all systems.

A failure to do so can lead to data inconsistencies and errors, which can have serious consequences for the business. For example, if the AI system updates the inventory levels in the enterprise resource planning system, but not in the product lifecycle management system, this can lead to discrepancies in the data and incorrect decisions being made.

Example of read and write paths

A shift supervisor checks the maintenance log in the computerised maintenance management system to confirm the schedule for the upcoming maintenance tasks. The AI system reads the data from the maintenance log and uses it to predict when the next maintenance task is due. However, when the AI system updates the maintenance schedule, it must write the data back to the computerised maintenance management system, ensuring that the data is consistent across all systems.

Reconciling definitions

Different systems often have different definitions for the same data element, which can make integration with AI systems challenging. For example, the definition of a "product" may vary between the enterprise resource planning system and the product lifecycle management system. Reconciling these definitions is critical to ensuring that the data is consistent and accurate across all systems.

This requires a detailed understanding of the data models and definitions used in each system, as well as the ability to map these definitions to a common standard. This can be a time-consuming and complex task, but it is essential to ensuring that the AI system is able to provide accurate and reliable insights.

Most people assume that reconciling definitions is a simple task, but it is often more complicated than it seems. The complexity of the data models and definitions, combined with the need to ensure consistency across all systems, makes it a challenging task that requires careful planning and execution.

Latency budgets and precomputation

Latency budgets refer to the amount of time that the AI system has to respond to a query or request. In the textiles and apparel industry, latency budgets can be tight, particularly when it comes to responding to changes in demand or adjusting pricing and inventory levels. To meet these latency budgets, it is often necessary to precompute certain data elements or queries, so that the AI system can provide fast and accurate responses.

Precomputation involves calculating the results of certain queries or data elements in advance, so that they can be quickly retrieved when needed. This can help to reduce the latency of the AI system and improve its overall performance. However, it requires careful planning and management, as the precomputed data must be kept up-to-date and consistent with the underlying systems of record.

For more information on how to implement precomputation and manage latency budgets, please visit our resources page.

Limitations and wrong choices

While AI systems can provide significant benefits for textiles and apparel companies, they are not a solution for every problem. In some cases, the complexity of the data or the systems of record may make it difficult or impossible to integrate with an AI system. In other cases, the latency budgets may be too tight, or the data may be too inconsistent or inaccurate.

In these cases, it may be necessary to consider alternative solutions, such as manual data analysis or reporting. It is also important to carefully evaluate the potential benefits and limitations of an AI system before implementing it, to ensure that it is the right choice for the business. For more information on how to evaluate the potential benefits and limitations of an AI system, please contact us.