• Written by Azhar Aliyar, Director

Can we really trust our AI system to answer questions about stock levels and sales trends when the data it's trained on is distorted by promotional activities and write-offs due to short shelf life? This is a question that often comes up in meetings with FMCG companies looking to integrate AI systems into their operations.

FMCG complexities

The fast-moving consumer goods industry presents unique challenges when it comes to integrating AI systems. Promotional activities, such as discounts and buy-one-get-one-free offers, can create distortions in historical sales data, making it difficult to identify genuine trends. Additionally, the short shelf life of many FMCG products means that stock levels and inventory management are critical, and any errors or inaccuracies can have significant consequences. Distributor visibility is also a challenge, as companies often rely on third-party distributors to get their products to market, which can make it difficult to track stock levels and sales in real-time.

These complexities can make it difficult to integrate AI systems, as they require high-quality, accurate data to function effectively. If the data is distorted or incomplete, the AI system's outputs will be compromised, which can lead to poor decision-making and lost sales.

Read and write paths

When integrating AI systems into FMCG operations, it's essential to consider both the read and write paths. The read path refers to how the AI system accesses the data it needs to function, while the write path refers to how the AI system updates or modifies the data. While the read path is relatively straightforward, the write path requires more care, as it involves updating or modifying data in the systems of record. This can be a challenge, as it requires ensuring that the AI system's updates are accurate and consistent with the company's existing data management processes.

A good example of this is when a category manager uses the AI system to analyze sales trends and identify opportunities to optimize product placement. The AI system may need to update the product catalog or inventory management system with new information, which requires careful consideration of the write path to ensure that the updates are accurate and consistent.

Reconciling definitions

One of the biggest challenges in integrating AI systems into FMCG operations is reconciling definitions that differ between systems. For example, the definition of a "sale" may differ between the sales database and the inventory management system. The sales database may define a sale as the moment a product is shipped to a customer, while the inventory management system may define a sale as the moment a product is received by a customer. These differences in definition can create inconsistencies and errors, which can compromise the accuracy of the AI system's outputs.

To overcome this challenge, companies need to invest time and effort into reconciling these definitions and ensuring that they are consistent across all systems. This may involve creating a data governance framework that outlines the definitions and standards for data management, as well as implementing data validation and quality control processes to ensure that the data is accurate and consistent.

Latency budgets and precomputation

FMCG companies often have high-volume, low-latency requirements, which can make it challenging to integrate AI systems. The AI system needs to be able to process large amounts of data in real-time, while also ensuring that the outputs are accurate and consistent. To overcome this challenge, companies can use latency budgets to prioritize the processing of critical data, such as sales trends and inventory levels. They can also use precomputation to prepare data in advance, which can help to reduce the latency and improve the responsiveness of the AI system.

For example, a company may use precomputation to prepare sales forecasts and inventory levels for the next day, which can help to improve the responsiveness of the AI system and reduce the latency. This can be particularly useful in FMCG, where sales trends and inventory levels can change rapidly, and companies need to be able to respond quickly to changes in the market. For more information on how to implement latency budgets and precomputation, please visit our website.

Limitations and wrong choices

While AI systems can be a powerful tool for FMCG companies, they are not a silver bullet, and there are limitations to their use. For example, AI systems are not suitable for tasks that require human judgment or creativity, such as negotiating with suppliers or developing new product lines. Additionally, AI systems may not be the best choice for companies with limited data management capabilities or inadequate data quality.

In these cases, companies may need to invest in data management and quality control processes before implementing an AI system. They may also need to consider alternative solutions, such as business intelligence or data analytics platforms, which can provide similar insights and benefits without the need for AI. If you're considering implementing an AI system and want to discuss your options, please contact us to speak with one of our experts.