Build a Strong Foundation for Your AI Journey

You are starting on your AI journey, but do you know what is the most important thing to do to build a strong foundation for AI to work?

In the WMS world, we typically only want to keep active data for looking up past information. We setup the archive data, the retention policies. We purge data from the system on a regular basis in order to keep the database clean and quick to query.

But for prediction, you need to think about different set of data, let's think about a few questions that the AI is going to ask:

• How did the demand change over the past 3 years?
• What did you sell in December last year and was that influenced by a special event?
• What was your return rate in that same period, and was there a spike due to a quality issue?
• How do your customers behave? Do they tend to re-order every 90 days, or is the pattern more random?

Without this type of information, the AI will not be able to predict with accuracy, and your investment in AI models will be limited in their effectiveness. For a meaningful answer, AI typically needs 1 to 3 years of demand data to detect seasonality and trends, 6 to 12 months of behavioral data to establish baselines (e.g., return rates, reorder frequency), and clearly labeled events (e.g., promotions, stockouts, supplier delays) to separate anomalies from real patterns.

Organizations taking up this AI journey for the first time should consider this important aspect: Starting the AI initiative does not have to start from building the AI models. The most impactful first step is to audit, structure, and begin collecting the foundational data. Think of it as training camp before the season begins — without it, even the best team (or model) will underperform.