Rohit Laila is a seasoned veteran who has spent decades navigating the intricate web of global logistics, from the floor of the warehouse to the heights of digital innovation. His deep-seated passion for technology is matched only by his pragmatism regarding its implementation, particularly in an era where artificial intelligence is often viewed as a panacea for every operational headache. In this discussion, we delve into the structural failures that lead to “pilot purgatory” and the essential strategies for scaling AI, focusing on the critical need for data harmonization and enrichment. We also explore why transparency is the non-negotiable key to fostering human trust in automated systems, ensuring that technology serves to augment rather than replace the intuition of experienced professionals.
The industry often speaks about “pilot purgatory,” where AI projects thrive in isolation but fail when introduced to the wider organization. From your perspective, why does this transition remain so difficult for most companies?
The jump from a small, controlled pilot to a full-scale deployment is where most companies feel the floor fall out from under them. The primary culprit is the sheer fragmentation of data coming from dozens of disparate sources, ranging from ERP systems and warehouse platforms to IoT sensors and supplier portals. These systems are rarely interconnected, meaning each one is essentially screaming in a different language with its own unique classification and taxonomy. When an AI model is fed this inconsistent noise, it loses the accuracy needed for high-stakes, enterprise-wide decision-making. The failure isn’t due to the complexity of the math, but a fundamental disconnection from a unified operational reality.
How do discrepancies in how different teams define key metrics, such as “promotional lift,” impact the reliability of AI models and the decisions based on them?
When you have a company where the sales, trade, and supply chain teams are all operating under different definitions of the same metric, you are setting your AI up for a total collapse. If these teams are feeding the model inconsistent versions of the truth, the resulting insights will be discrepant and, frankly, unusable for any real-world application. Without a rigorous process of harmonization to ensure all systems describe products and units consistently, even the most advanced AI is just comparing apples to oranges. Success requires building a cohesive library of truth rather than treating servers as cluttered storage units for mismatched files.
You have emphasized that internal data alone is often insufficient for a high-performing supply chain. How does marrying internal information with external context change the effectiveness of AI recommendations?
Internal data only tells you what happened within your own four walls, but the modern supply chain lives and breathes in a volatile global marketplace. By marrying internal figures with external context—like market conditions, competitive intelligence, and broad economic indicators—we give the AI the perspective it needs to provide truly actionable advice. This enrichment allows the system to understand the “why” behind a sudden dip or spike in demand, rather than just reporting the numbers in a vacuum. It turns a reactive tool into a predictive engine that can navigate the nuances of a shifting economy with much greater precision.
There is often a “black box” problem with AI where users don’t understand how a conclusion was reached. Why is explainability so critical for demand planning and organizational trust?
Black box models are dangerous because they undermine the confidence of the very people who need to use them, which ultimately increases organizational risk. A demand planning executive needs to know if a forecast was driven by a genuine shift in consumer behavior or just a seasonal trend that happens every year. Demanding traceability allows users at every level to question and validate recommendations instead of blindly following a machine’s output. This creates an accountable feedback loop where the system gets smarter over time while augmenting, rather than replacing, the invaluable expertise of our human teams.
Once a company successfully establishes these data foundations, what tangible transformations have you seen in their day-to-day decision-making cycles?
When the foundation is right, the impact on the business is breathtaking; we see decision cycles that once dragged on for weeks suddenly shrink down to mere minutes. Because teams are finally working from a single source of truth, the cross-functional friction and endless meetings to “align on numbers” simply evaporate. This consistency allows the entire organization to adapt quickly to market disruptions because the underlying data pipelines are traceable and reliable. It provides a level of agility that is impossible to achieve when your teams are still debating which spreadsheet contains the correct information.
What is your forecast for the future of AI in the supply chain industry?
I expect the gap between leaders and laggards will soon be defined by data infrastructure rather than the AI software itself. We are moving away from the era of experimental tinkering and toward a future where “accountable systems” are the gold standard for global logistics. In the coming years, winners will prioritize building a robust foundation of truth that can withstand any market shock, ensuring AI is a seamless extension of human decision-making. Those who continue to experiment on fragmented data will find themselves left behind by those who invested in the infrastructure of lasting value.
